E Shilton 383

Insuring Inequality: Sex-Based

Mortality Tables and Women’s

Retirement Income

Elizabeth Shilton

In Canada, retired women have considerably less financial resources than retired men. This gender gap is contributed to by the insurance industry’s use of sex-based mortality tables (SBMTs) in pricing annuities—one of the few remaining instances of direct sex discrimination still lawful in Canada.

SBMTs indicate that, statistically, women live longer than men, so insurers argue that it is “fair discrimination” to charge women more for annuities than men. This notion of fairness emphasizes individual rights—lower-risk individuals are protected from subsidizing higher-risk individuals. In the author’s view, this is a perverse social policy. Welfare risks should be addressed collectively, because an individual focus leaves women to their own unequal resources.

However, the author focuses not on market logic but on the legality of using SBMTs, which was indirectly established in the Zurich Insurance case in 1992. She argues that the development of reliable pricing tools and the subsequent evolution of Canadian human rights jurisprudence present a strong case against SBMTs. The provisions in human rights and insurance legislation allowing the use of SBMTs are also unlikely to hold up against the constitutional guarantee to equality under section 15 of the Canadian Charter of Rights and Freedoms. This position is supported by the 2011 decision of the European Court of Justice in Test-Achats, which removed an insurer’s immunity to price annuities based on SBMTs.

Introduction I. “Fair Discrimination”: The Evolution of SBMTs in Annuity Markets II. How SBMTs Affect Retirement Income in Canada III. Legal Rules, SBMTs and Employment Pension Equality: The US and

European Case Law A. “Fair Discrimination” Confronts Individual Rights: The US Cases B. Sex-Based Pension Distinctions and “Equal Pay”: The European Court

of Justice IV. Taking on the Insurance Industry: Association belge des Consommateurs

Test-Achats ASBL V. SBMTs in Canada

I am grateful to Bernie Adell and Karen Schucher, who provided very helpful

comments on an earlier draft of this paper. Thanks also to David Short and Shiraz Bhirmal who generously provided useful actuarial background.

384 (2012) 37:2 Queen’s LJ

A. The Statutory Framework B. “Reasonable and Bona Fide” Insurance Practices C. SBMTs and Contemporary Canadian Legal Equality Standards

Conclusion

Introduction

Like most developed countries, Canada is in the midst of an important policy debate about appropriate responses to the aging of its population. A central theme in this debate is the role of the state in addressing retirement income needs. By international standards, current Canadian retirement income arrangements leave an unusually large share of pension provision to the market, and employment-based pension plans do much of the heavy lifting.1 With employment pension coverage on a steady downward slide from its peak in the 1970s, the gap between state provision and retirement income needs has become a matter of substantial public concern.2

At the core of the policy debate is whether the risks of welfare loss posed by this gap are social risks or individual risks. Proponents of the “social risk” theory see the problem as best addressed by the social insurance mechanisms of the state. They look for solutions to the retirement income deficit in the expansion of state systems like the Canada/Quebec Pension Plan (CPP/QPP) systems which can address retirement risks collectively by pooling and distributing them across a broad social

1. British Columbia, Steering Committee of Ministers on Pension Coverage and

Retirement Income Adequacy, Options for Increasing Pension Coverage Among Private Sector Workers in Canada (Victoria, BC: Ministry of Finance, 2010) at 68. 2. Almost three-quarters of working-age Canadians do not have employment

pension plans. See Bob Baldwin, “Determinants of the Evolution of Workplace Pension Plans in Canada”, Caledon Institute of Social Policy (2007) at 14–19, online: <http://www.caledoninst.org>. For a discussion of the implications of the decline in employment pension coverage, see Baldwin Consulting, Research Study on the Canadian Retirement Income System by Bob Baldwin (Oshawa, Ont: Ministry of Finance, 2009), online: <http://www.fin.gov.on.ca/en/consultations/pension/dec09report.pdf>;

Keith Horner, “A New Pension Plan for Canadians: Assessing the Options”, Institute for Research on Public Policy (2011), online: <http://www.irpp.org>.

E Shilton 385

spectrum as well as across generations,3 leaving retirees much less vulnerable to individual misfortune and to the vagaries of financial markets. Proponents of the “individual risk” theory look to the market; as they see it, we need less state in the retirement income system, not more state. They argue that solutions lie in broader use of savings and investment vehicles to defer pre-retirement consumption in order to meet an individual’s retirement income requirements. They maintain that twentieth century welfare state-centred approaches to pension provision are unsustainable in a modern globalized economy, and the current regulatory burdens must be lightened to enable markets to respond creatively to evolving needs.4

Proponents of markets seek to frame the issue simply as a debate about means. The message is that the end—retirement income security for Canadians—is an undifferentiated commodity that either state or market can deliver. Markets can do the job

3. See e.g. Monica Townson, Options for Pension Reform: Expanding the Canada

Pension Plan, Policy Brief (Ottawa: Canadian Centre for Policy Alternatives, 2010); Canadian Labour Congress, Retirement Security for Everyone: Get the Job Done, online: Canadian Labour Congress <http://www.canadianlabour.ca>. 4. Key international organizations such as the World Bank and the Organisation

for Economic Co-operation and Development (OECD) promote a smaller role for national governments and a larger role for the private sector in pension provision. See e.g. The World Bank, Averting the Old Age Crisis: Policies to Protect the Old and Promote Growth, Policy Research Report (New York: Oxford University Press, 1994); The World Bank, Protecting the Old and Promoting Growth: A Defense of Averting the Old Age Crisis, Working Paper No 1570 (Washington, DC: Policy Research Department, 1996); OECD, Maintaining Prosperity in an Ageing Society, (Paris: OECD, 1998); OECD, Reforms for an Ageing Society: Social Issues (Paris: OECD, 2000); OECD, Closing the Pensions Gap: The Role of Private Pensions, Policy Brief (Paris: OECD, 2007) (role of the private sector in pension provision is expanding internationally). In Canada, the insurance industry has been an enthusiastic proponent of private sector market-based solutions to the retirement income deficit. See e.g. Sun Life Financial Canada, News Release, “Proposal to Reform Pension System Achieves Public Interest Criteria: Sun Life” (10 December 2009), online: <http://www.sunlife.com>; Bill Kyle, “The Strength of CAPs in Canada’s Retirement Market: How Collaborative Reform Can Lead to a Stronger Pension System for Canadians” (e-presentation delivered at the Summit on the Future of Pensions, April 2010), online: The Conference Board of Canada <http://www.conferenceboard.ca>.

386 (2012) 37:2 Queen’s LJ

more cheaply and efficiently than any expanded public system, making market-based retirement instruments the obvious choice for cash-strapped governments and heavily burdened taxpayers. Their arguments suggest that all Canadians will benefit equally from the choice of market over state provision, and will all be equally disadvantaged if policy makers are foolish or misguided enough to make the opposite choice. This simplistic narrative obscures vital connections between market logic and legal rules, and between legal rules and distributive outcomes. It is important for policy makers to understand these connections and to take them into account in the current pension reform process. Otherwise, there will be serious negative consequences for groups of Canadians already disadvantaged under our current retirement income arrangements. One of these groups is women.

The history of sex-based mortality tables (SBMTs) in annuity markets provides a useful case study, both on the role played by market logic in shaping legal rules and the role played by legal rules in shaping distributive outcomes. Canadians who seek the security of a guaranteed retirement income stream—the type of security offered by government pension plans like the CPP/QPP and by some private pension plans, particularly of the defined benefit variety—are advised by market proponents that there is a market solution available to meet their needs: the purchase of an annuity. Women who heed this advice and take their retirement nest eggs to market find themselves facing annuity prices some 10 per cent to 15 per cent higher than prices for men with comparable levels of retirement savings.5 This pricing structure is

5. In their World Bank study, Estelle James & Dimitri Vittas show an average

sex-based annuity price differential in Canada of approximately 10% to 15%. “Annuities Markets in Comparative Perspective: Do Consumers Get Their Money’s Worth?” World Bank Policy Research Paper No 2493 (paper delivered at the World Bank Conference on New Ideas About Old Age Security, September 1999) at 28–29. This is consistent with Canadian annuity quotes available on the Morningstar website, which show sex-based rate differentials ranging from about 9% to 15% for single life annuities purchased at age 70 (figures for age 65 are not listed). See Morningstar Canada, “A Summary of the Highest and Lowest Consumer Credit and Investment Rates Across Canada”, online: <http://www.morningstar.ca> (in general, the differential increases as purchasers get older).

E Shilton 387

an artifact of the insurance industry and results from using a pricing tool—SBMTs—which reflects women’s statistical longevity. Since the average woman lives longer than the average man and will therefore collect on her annuity longer, the industry insists that she pay more up front for that annuity. SBMTs are one of the few remaining instances of direct sex discrimination still legally tolerated in Canada. They are vigorously defended by the insurance industry as necessary to the efficient functioning of insurance markets.

The legal framework applicable to the insurance industry has to date permitted SBMTs to flourish in financial markets where individual annuities are sold.6 Very different legal rules govern employment pension plans. These rules do not permit sex-based differences in mortality to affect employee contribution rates or benefit levels. Likewise, SBMTs are not permitted to play any role in setting contribution or benefit levels under the CPP/QPP. These differences in the rules governing retirement income instruments raise the stakes for women in the current retirement income policy debate. If SBMTs depress women’s retirement income—and we will see in this paper that they do—an expanded role for the retirement income instruments that use them will have negative consequences for women’s retirement income. It is important to examine the legal issues raised by SBMTs in order to understand the gender impact of the retirement income policy choices now facing Canada. Such an examination is made even more timely by the 2011 holding of the European Court of Justice (ECJ), in the case of Test-Achats,7 that the use of SBMTs in insurance markets contravenes European equality guarantees.

In this paper, I look at the legal questions of whether SBMTs contravene Canadian human rights laws and constitutional equality guarantees. I also have more ambitious objectives. I place my analysis within a context that examines the links between the

6. The use of SBMTs is by no means mandatory in open annuity markets, but it

appears that Canadian insurers use SBMTs wherever they can legally do so. See email from the Canadian Life and Health Insurance Association to Elizabeth Shilton (17 May 2011) [on file with the author]. 7. Association belge des Consommateurs Test-Achats ASBL v Conseil des ministres,

C-236/09, [2011] 2 CMLR 994 [Test-Achats].

388 (2012) 37:2 Queen’s LJ

market logic that generated SBMTs and the legal rules that govern their use. In doing so, I identify broader lessons from the legal history of SBMTs that may inform the pension reform process in Canada. In Part I, I explore the connection between SBMTs and the insurance concept of “fair discrimination” or “efficient discrimination”, highlighting the perverse relationship between actual retirement income needs and the market model of risk distribution. In Part II, I examine the contribution of SBMTs to the overall retirement income gender gap in Canada. In Part III, I take an overview of how US and European courts have dealt with the use of SBMTs within employment pension plans, pitting the market’s “efficient discrimination” perspective against an anti-discrimination perspective rooted in individualistic rights theory. In Part IV, I explore the legal immunity enjoyed by the insurance industry from the restrictions applied to SBMTs within employment pension plans, and the withdrawal of that immunity in Europe as a consequence of the ECJ’s Test-Achats decision. In Part V, I analyze the current Canadian legal framework that governs the use of SBMTs, and assess the likelihood of a successful challenge to that framework in the wake of Test-Achats. I conclude that SBMTs are incompatible with Canada’s human rights and constitutional commitments to sex equality. In closing, I argue that this case study offers evidence in support of some broader propositions which should be taken into account in the policy process: that legal rules have important distributive consequences, that rules shaped by the market logic of risk distribution tend toward substantively unequal outcomes, and that retirement income instruments built on social insurance principles of broad risk pooling are more likely to promote retirement income equality than instruments which conceive of retirement income as a problem to be solved on an individual basis.8

8. While my analysis in this paper is confined to the retirement income context,

the debate about social insurance versus market insurance in promoting welfare is a much broader one. See Regina Austin, “The Insurance Classification Controversy” (1983) 131:3 U Pa L Rev 517; Jonathan Simon, “The Ideological Effects of Actuarial Practices” (1988) 22:4 Law & Soc’y Rev 771; Trudo

E Shilton 389

I. “Fair Discrimination”: The Evolution of SBMTs in Annuity Markets

In order to understand the impact of sex-based annuity pricing on retirement income, it is necessary to understand what annuities are and how they work. An annuity is a series of fixed payments made over a period of time; if payable for life, it is a “life annuity”.9 Life annuities (I will simply call them annuities) may be payable under the terms of an employment pension plan (in which case they are normally called “pensions”), or under the terms of a free-standing contract between an individual and an insurance company. Recipients of annuities see them as a retirement income stream. From the perspective of annuity providers, however, annuity contracts are properly characterized as insurance policies—“reverse life insurance” or “longevity insurance”—for which purchasers pay a lump sum in return for the vendor’s promise to make periodic payments until death.10 Consistent with this insurance perspective, the Canadian annuity market is almost entirely occupied by life insurance companies.11

Because annuity contracts are both contingent12 and open-ended, they pose special challenges for profitable pricing. Profits depend largely on the insurer’s ability to predict how long the

Lemmens, “Selective Justice, Genetic Discrimination, and Insurance: Should We Single Out Genes in Our Laws?” (2000) 45:2 McGill LJ 347. 9. James M Poterba, “Annuity Markets” in Gordon L Clark, Alicia H Munnell

& J Michael Orszag, eds, The Oxford Handbook of Pensions and Retirement Income (Oxford: Oxford Universtiy Press, 2006) 562. 10. This is the basic model, although both life insurance policies and annuities

can be structured in a wide variety of ways. For example, they may provide payouts only until a certain age rather than for life. See E Philip Davis, Issues in the Regulation of Annuity Markets, Working Paper 26/02 (Moncalieri, Italy: Centre for Research on Pensions and Welfare Policies, 2002) at 3–4. 11. Some fraternal benefit societies still provide annuities. The federal

government sold its own annuities from 1908 until the 1960s, when they were discontinued due to pressure from the insurance industry. See Kenneth Bryden, Old Age Pensions and Policy-Making in Canada (Montreal: McGill-Queen’s University Press, 1974) at 51–59. 12. The technical label for contracts where performance depends on a

contingency is “aleatory”.

390 (2012) 37:2 Queen’s LJ

annuitant will live. In the case of life insurance, long-lived purchasers present the lowest risk for insurers. In the case of annuities, by contrast, purchasers with longer life expectancy pose a high financial risk to insurers since the annuity will likely have to be paid out longer. Insurers use statistical techniques developed by actuaries to predict longevity.13 Despite the fact that these techniques are most accurate over large populations, insurers in voluntary insurance markets resist universal pricing, preferring to classify purchasers into “different risk pools” defined by characteristics correlated positively or negatively with longevity. Insurance contracts are then priced according to the risk level of each pool. Differential risk pools protect insurers against “adverse selection”—the possibility that clients who estimate their life span as shorter than average will not buy annuities because they see them as too expensive, leaving insurers with portfolios of long-lived (i.e. high-risk) clients who have secured a bargain by insuring themselves at average risk rates.14 Pricing based on different risk pools makes insurance profits more reliable, and expands the potential client base by allowing insurers to service both extremely high-risk and extremely low-risk clients at profitable rates.15

The risk assessment process is frequently labeled “actuarial science”. There undoubtedly are scientific aspects to the methodology, but science does not dictate the boundaries of risk pools; there is considerable room for the exercise of both actuarial and business judgment.16 Nowhere is this more apparent than in the history of sex-based insurance classifications. Insurers and their advocates give SBMTs a high ontological status,

13. See Timothy Alborn, Regulated Lives: Life Insurance and British Society, 1800-

1914 (Toronto: University of Toronto Press, 2009). 14. See Wouter PJ Wils, “Insurance Risk Classifications in the EC: Regulatory

Outlook” (1994) 14:3 Oxford J Legal Stud 449 at 449–51; Thomas Flanagan, “Insurance, Human Rights and Equality Rights in Canada: When is Discrimination ‘Reasonable’?” (1985) 18:4 Can J Pol Sc 715 at 726–72. 15. See Kenneth S Abrahams, “Efficiency and Fairness in Insurance Risk

Classification” (1985) 71:3 Va L Rev 403 at 408. 16. See Leah Wortham, “The Economics of Insurance Classification: The Sound

of One Invisible Hand Clapping” (1986) 47:4 Ohio St LJ 835 at 845–47 [Wortham, “Economics”].

E Shilton 391

characterizing them as reflecting obvious and incontrovertible truths about the world.17 Timothy Alborn’s account of the evolution of sex-based risk classification, however, makes it clear that there was no eureka moment at which actuaries discovered that males and females had distinctly different life spans. Alborn’s tale is one of historical contingency, marketing strategy and gender bias. As he explains it, nineteenth century insurers based their standard premium schedules on “life tables” derived from the vital statistics of the “healthy male”—the family breadwinner—to whom life insurance was primarily marketed.18 Women were often denied insurance or were subject to arbitrary surcharges. When life insurers did offer women the same rates as men, Alborn points out that women still paid “more at older ages than their life expectancy warranted”.19 Until the passage of the Married Women’s Property Acts in the latter part of the nineteenth century, married women could not insure their own lives because they had limited capacity to contract.20 Only when women emerged as potential consumers of life insurance in their own right did insurers begin to use sex systematically as a risk

17. See e.g. Spencer L Kimball, “Reverse Sex Discrimination: Manhart” (1979)

4:1 ABF Research J 83 at 96. Kimball argued that states which prohibit the use of SBMTs are requiring insurers “to ignore reality and to act in accordance with the fiction that men and women do not differ”, and describes unisex mortality tables as a “technical device for concealing the facts”. Ibid at 96, 133. 18. Alborn, supra note 13 at 103. 19. Ibid at 9. See also ibid at 103, 116–121. Efforts in the early 19th century to

acknowledge female longevity through lower life insurance premiums were catastrophic for the few companies which attempted the experiment. Insuring the lives of women was exceptional in that era, and the laws of adverse selection ensured that the clients enticed into the female market were insuring women in their child-bearing years whose mortality was considerably higher than that of men of similar age (ibid at 117). See also George J Benston, “The Economics of Gender Discrimination in Employee Fringe Benefits: Manhart Revisited” (1982) 49:2 U Chicago L Rev 489 at 530 [Benston, “The Economics of Gender Discrimination”]. 20. See Constance Backhouse, “Married Women’s Property Law in Nineteenth-

Century Canada” (1988) 6:2 LHR 211. C.f. Alborn, supra note 13 (stating that most 19th century insurance policies taken out on women’s lives were purchased by husbands insuring their wives’ expectation of an inheritance, an expectation that would not be consummated if their wives predeceased the testator).

392 (2012) 37:2 Queen’s LJ

factor, offering women lower rates to reflect their higher life expectancy.21 Because women’s superior longevity posed a high financial risk, insurers were not so anxious to attract women as annuity clients. They frequently attached a premium to female annuity rates much larger than the discount offered to women for life insurance.22 Sex-based insurance rates, then, were a reflection of what gendered markets would bear.

Although sex has proved to be a durable risk classifier, biology is not destiny here. Sex is widely if not universally23 acknowledged as only a proxy for gendered lifestyle and environmental factors such as smoking, alcohol consumption and workplace stress, all of which are more plausible indicators of life expectancy than sex. As a proxy, however, sex is perceived as reasonably efficient because it easily meets key criteria insurers rely on to classify risk—availability of information, cheapness and verifiability.24 Its stability and reliability, although just as important, are more debatable, and have indeed been vigorously debated.25 Nevertheless, insurers remain firmly committed to

21. Leah Wortham, “Insurance Classification: Too Important to be Left to the

Actuaries” (1986) 19:2 U Mich JL Ref 349 at 375 [Wortham, “Insurance Classification”]. 22. A popular approach to adjusting mortality tables was to apply the “set-back

method” to male mortality tables, treating women as if they were younger men. For life insurance, the set-back was typically three years; for annuities, it was the six years actually reflected in the statistics. See ibid at 376–77; Kimball, supra note 17 at 109. Insurers invoked adverse selection to justify this distinction. 23. Kimball is out of the mainstream in claiming a biological link. He argues:

“to analogize sex and race in discussing mortality is fallacious. Use of race is, for estimation of mortality, wholly indefensible on factual grounds. Use of sex is not only defensible but it reflects as well both common sense and the available facts”. Ibid at 113. 24. Flanagan, supra note 14 at 726–27. See also Abrahams, supra note 15 at 410–

20. 25. Much of what has become known as the Brilmayer-Benston debate over

SBMTs turned on the factual question of whether sex is a reliable and stable classifier. See Lea Brilmayer et al, “Sex Discrimination in Employer-Sponsored Insurance Plans: A Legal and Demographic Analysis” (1980) 47:3 U Chicago L Rev 505 at 512 [Brilmayer et al, “Sex Discrimination”]; Lea Brilmayer, Douglas Laycock & Teresa A Sullivan, “The Efficient Use of Group Averages as Non-Discrimination: A Rejoinder to Professor Benston” (1983) 50:1 U Chicago L Rev 222; Benston, “The Economics of Gender Discrimination”, supra note 19; and

E Shilton 393

SBMTs, arguing that they are statistically valid and perhaps even essential economic tools for risk management.26

In the terminology used by insurers, the key issue is whether SBMTs are “fair” classifiers, an industry term of art widely reflected in statutory standards for insurance rate-setting.27 Insurance “fairness” is not an ethical concept; it is judged almost entirely by statistical measures. If the statistics that differentiate

George J Benston, “Discrimination and Economic Efficiency in Employee Fringe Benefits: A Clarification of Issues and a Response to Professors Brilmayer, Laycock and Sullivan” (1983) 50:1 U Chicago L Rev 250 [Benston, “Discrimination and Efficiency”]. Whether or not they were ever stable and reliable, statistical gender differences in mortality are certainly shrinking. While earlier statistics reflected six-year differences, the most recent table adopted by the actuarial profession in Canada, the UP-94 Table, shows a sex difference in 1994 of 3.8 years at age 65, projected to shrink to 2.4 years by 2024. See Society of Actuaries UP-94 Task Force, “The 1994 Uninsured Pensioner Mortality Table” (1995) 47 Transactions of the Society of Actuaries 819, Table 7 at 838–39. 26. TIAA-CREF took the position in litigation that SBMTs were a “business

necessity”. See Sydney J Key, “Sex-Based Pension Plans in Perspective: City of Los Angeles, Department of Water and Power v. Manhart” (1979) 2 Harv Women’s LJ 1 at 4. See also City of Los Angeles Department of Water and Power v Manhart, 435 US 702, 98 S CT 1370 (1978), Burger CJ, dissenting (describing sex as “the only objective feature upon which an employer—or anyone else, including insurance companies—may reliably base a cost differential for the ‘risk’ being insured” at 727). 27. See Unfair or Deceptive Acts or Practices, O Reg 7/00, s 1 under Insurance Act,

RSO 1990, c I.8, s 438. “Unfair and deceptive practices” are described as including:

2. Any unfair discrimination between individuals of the same class and of the same expectation of life, in the amount or payment or return of premiums, or rates charged for contracts of life insurance or annuity contracts, or in the dividends or other benefits payable on such contracts or in the terms and conditions of such contracts.

3. Any unfair discrimination in any rate or schedule of rates between risks in Ontario of essentially the same physical hazards in the same territorial classification.

“Fairness” is also the general US regulatory standard. See Robert H Jerry II & Kyle B Mansfield, “Justifying Unisex Insurance: Another Perspective” (1985) 34:2 Am U L Rev 329 at 336; Wortham, “Insurance Classification”, supra note 21 at 387–93.

394 (2012) 37:2 Queen’s LJ

risk categories are valid by actuarial standards, the categories they create are “fair” as measured by the insurer’s yardstick. A “fair” risk category avoids the evil of “subsidies”, through which low-risk clients pay more than a “fair” price in relation to their risk and higher-risk clients pay less. Paradoxically, in view of the group nature of insurance, “insurance fairness” is essentially an individualistic concept, based on the precept that individuals should pay their own way to the extent compatible with the nature of insurance products. Just what that extent is will be determined by the insurance industry through the mechanism of risk classification, a determination that will ultimately be ratified by the competitive market.28

The discrimination created by insurance risk categories is claimed to be “fair” or “efficient discrimination”, normatively distinguishable from the invidious discrimination targeted by civil rights/human rights statutes.29 This concept of fairness is not grounded in any substantive idea of equality. It simply requires that risk groups be reasonably homogeneous, and that “likes” be treated alike. But the insurance industry certainly makes equality claims for “fair discrimination”. Indeed, it goes so far as to claim that it is unisex mortality tables, and not SBMTs, that invidiously discriminate on the basis of sex. This argument is painstakingly laid out in a 1979 article by Spencer Kimball.30 That article is an

28. Benston argues that the competitive marketplace can be trusted to inhibit

insurers from practicing invidious sex discrimination, since this would be inefficient. Benston, “The Economics of Gender Discrimination”, supra note 19 at 529 (insisting that “[t]here is no evidence indicating that the identification of gender as a determinant was related to any prevailing stereotypes or other prejudices” at 530). But see Alborn, supra note 13 (providing such evidence in his historical account to rebut Benston). 29. See e.g. Kimball, supra note 17; Benston, “The Economics of Gender

Discrimination”, supra note 19; Benston, “Discrimination and Efficiency”, supra note 25. See also Flanagan, supra note 14; David D McCarthy & John A Turner, “Risk Classification and Sex Discrimination in Pension Plans” (1993) 60:1 The Journal of Risk and Insurance 85. 30. Kimball, supra note 17. Kimball’s article might seem an obsolete target for

criticism were it not for the fact that the insurance industry continues to make these same arguments to annuity contracts on open financial markets. See also Yves Thiery & Caroline Van Schoubroeck, “Fairness and Equality in Insurance Classification” (2006) 31:2 The Geneva Papers on Risk and Insurance 190

E Shilton 395

extended critique of the 1978 US Supreme Court decision in Manhart,31 in which the Court held that differences in employee pension contributions based on sex violate Title VII of the Civil Rights Act, 1964.32 The core of Kimball’s argument is that in assessing whether or not employment pension plans deliver equal benefits, it is an analytic error to focus on whether men and women receive equal income streams from the plan; the only meaningful standard of equality is “present value” (also labelled “actuarial value”), an actuarial construct that can only be properly measured using mortality tables that take into account women’s superior longevity. If men and women receive equal income streams, women’s pensions have a higher actuarial value measured by SBMTs than those of similarly situated men, forcing men to “subsidize” women in violation of fundamental principles of insurance “fairness”.

The “equal actuarial value” argument rests on some important assumptions: first, that compensation theory dictates a comparative focus on employer contributions rather than on periodic benefits;33 and second, that sex-based actuarial values reflect a “real” equality in the “real” world that will be unfairly ignored if SBMTs cannot be used. This second assumption is a relatively easy target. As numerous critics have pointed out,34 its elementary flaw is that it is entirely circular, since it depends for its “reality” on legal rules that permit risk classification on the basis of sex.35 The first assumption is also contested. Those who

(discussing the industry lobby against the Council of Europe’s Directive 2004/113 regulating insurance risk categories, adopted in 2004 and addressed in Test-Achats, supra note 7). 31. Manhart, supra note 26 (discussed in more detail in Part III below). 32. Title VII of the Civil Rights Act, 42 USC § 2000e (1964). This federal statute

prohibits employment discrimination on grounds including sex, and is the functional equivalent of employment discrimination prohibitions in Canadian human rights codes. 33. Kimball, supra note 17 (noting that the employer’s cost is “the one and only

thing that is a clear equivalent or recompense for work done” at 98). 34. See e.g. Brilmayer et al, “Sex Discrimination”, supra note 25 at 512–14; Wils,

supra note 14 at 459. The US Supreme Court accepted the circularity critique, discussed in Part III below. 35. The argument rests also on two other assumptions which are fundamental to

Kimball’s equal pay thesis: first, that employer contributions are directly

396 (2012) 37:2 Queen’s LJ

argue that the proper focus of comparison is the pension income stream rather than employer contributions point to a competing compensation theory which characterizes pension schemes as implicit contracts for a continued income stream;36 pensions are deferred wages, and deferred wages (like current wages) should be governed by equal pay norms.37 Proponents of equal income streams also argue that even if there are stable and reliable differences between male and female longevity (and most of them do not concede that there are), statistical differences are not meaningful to actual pension recipients, since most women have lifespans comparable to most men.38 Ultimately, however, their primary critique of the “equal actuarial value” argument is a normative one; they argue that sex-based classifications, like race-based classifications, are invidious per se.39 The opponents of “fair discrimination” acknowledge, perhaps more frankly than its proponents do, that it is a divergence of views on appropriate

calibrated to wages and second, that overall benefits from pension plans are entirely dependent on longevity. Under modern funding rules, the first assumption is true only for defined contribution plans; it is often not true at all for defined benefit plans, where it is benefits that are normally pegged to wages, and employer contributions are highly variable. See Bernard L Adell, Pension Plan Surpluses and the Law: Finding a Path for Reform, Reprint Series No 75 (Kingston: Queen’s Industrial Relations Centre, 1988). The second assumption has been rendered largely obsolete by modern pension regulatory rules that mandate guaranteed death benefits and survivor benefits. These rules move away from longevity or the sole protection of employee financial interests, by ensuring that survivors of a short-lived employee reap at least some of the financial benefit from the plan that the employee did not live to enjoy. These points are not discussed in the academic debate around SBMTs. 36. See Richard A Ippolito, Pension Plans and Employee Performance: Evidence,

Analysis and Policy (Chicago: University of Chicago Press, 1997) at 17 (characterizing pensions as an implicit contract for the payment of a benefit indexed to the final wage). 37. This argument is discussed in detail in Merton C Bernstein & Lois G

Williams, “Title VII and the Problem of Sex Classifications in Pension Programs” (1974) 74:7 Colum L Rev 1203 at 1222–23 (although they do not use the now-common expression “deferred wages”, this is the essence of their compensation argument). 38. See generally supra note 25. 39. See Brilmayer et al, “Sex Discrimination”, supra note 25; Jerry & Mansfield,

supra note 27; Bernstein & Williams, supra note 37.

E Shilton 397

distributive outcomes, and not disagreements about logic or science or compensation theory, which lies at the core of the debate.40

The insurance concept of “actuarial equality” is highly abstract, even for the rare individual woman whose lifespan is at the statistical average. Annuitants are certainly concerned about the durability of their retirement income stream, and they value the hedge against longevity risk embedded in the actuarial equality concept.41 Of more immediate consequence, however, is the adequacy of that income stream; what matters from day-to-day is whether pension benefits are large enough to pay the bills. As an exercise in market logic, a system in which those with the longest prospective lifespan have to pay more to get pension benefits may be persuasive. As social policy, it appears simply perverse.

II. How SBMTs Affect Retirement Income in Canada

An approach that makes those most in need of protection pay the most for it would make distributive sense if there were a positive correlation between longevity and pre-retirement earning capacity—if those with the greatest need were also those with the greatest resources. In the case of gender, there is no such positive correlation. In fact, the reverse is notoriously true.42 The gender

40. See ibid; Wortham, “Insurance Classification”, supra note 21; Austin, supra

note 8; Simon, supra note 8. 41. Not all employment pension plans provide this hedge. Regulatory laws in

most Canadian jurisdictions do not require defined contribution plans to mandate that retiring employees annuitize their defined contribution accounts. Several Canadian provinces permit plan members with locked-in pension accounts to transfer significant amounts of their funds to regular (i.e. not locked-in) tax-sheltered investment accounts. See e.g. O Reg 239/09. Like other investment accounts, these accounts may be drained before the retiree dies, particularly if the investments do not do well. 42. See Statistics Canada, Women in Canada: A Gender-Based Statistical Report,

6th ed (Ottawa: Minister of Industry, 2011) (the most recent documentation of the economic inequality of women in Canada). See especially Cara Williams, “Economic Well-Being” in ibid. See generally supra note 5.

398 (2012) 37:2 Queen’s LJ

gap in retirement income in Canada currently stands at about 30 per cent in favour of men.43 The anatomy of this gap has not been well studied, and accessible data sources are not very satisfactory for that purpose.44 The question of what impact SBMTs currently have on the retirement income of Canadian women is therefore not easy to answer precisely. As I will shortly explain, however, the information available suggests that SBMTs cost Canadian women some $1.155 billion annually.

Retirement income in Canada comes from a variety of sources. Table 1, below, sorts federal government data on five categories of income sources.45

43. In 2009, average income for Canadian men 65 and older was $37 000; for

women, it was $25 700. “Income: Social Indicators”, Table 202-0407, online: Statistics Canada <http://www.statcan.gc.ca/pub/11-008-x/2006007/t/4169291-eng.htm>. 44. Availability of accurate data has been a chronic problem for Canadian

pension scholars for at least a century. See Expert Commission on Pensions, A Fine Balance: Safe Pensions, Affordable Plans, Fair Rules (Ontario: Queen’s Printer, 2008) at 29. 45. The data in Table 1 cannot be added together to provide total average

incomes (or to assess what portion of average income comes from these individual sources) because it reports only on persons who actually receive income from those sources. A significant number of elderly Canadians have no income from many of these sources, which is why the overall average set out in note 43 above is lower than would be indicated by average income levels from these sources.

E Shilton 399

Table 1: Average Income for Canadian Seniors, by Gender and Income Source (in Canadian Dollars)

OAS/GIS CPP/QPP Investments RPP/RRSP46 Earned Income

M F M F M F M F M F

2005 6 400 7 100 6 800 5 300 6 100 4 400 18 200 11 000 13 100 7 600

2007 6 700 7 400 7 000 5 500 6 600 5 000 16 200 13 900 14 400 9 200

2009 6 900 7 700 7 400 5 900 6 700 4 900 20 200 13 300 15 800 10 700Source Data: Indicators of Well-Being in Canada Retirement Income (Human Resources and Skills Development Canada) <http://www4.hrsdc.gc.ca/.3ndic.1t.4r@-eng.jsp?iid-27>, last accessed 25 June, 2011, supplemented by 2009 data released by the Income Statistics Division, Statistics Canada in June 2011.

46. Table 1 shows significant fluctuations in the RPP/RRSP category between 2005 and 2009, almost certainly accounted for primarily by a 2007 income tax rule change permitting spouses to split pension income for tax purposes. This change produced a dramatic net transfer from males to females in the RPP/RRSP column in 2007, which appears to have been reversed somewhat by 2009.

400 (2012) 37:2 Queen’s LJ

The only category in which the gender gap favours women is

the Old Age Security/Guaranteed Income Supplement (OAS/GIS) category. This category consists of a universal public pension clawed back at higher income levels, combined with a supplement for Canadians with the lowest incomes; therefore, the fact that women have higher levels of OAS/GIS in this category actually reflects their greater poverty. In the CPP/QPP column, men do better than women, but by a margin smaller than the overall gender gap, reflecting design features in this public plan that assign pension credit to women’s work patterns and minimize penalties for women’s unequal share of family work more effectively than private plans.47 There is no separate column for annuity income, which may fall into either the Registered Pension Plan/Registered Retirement Savings Plan (RPP/RRSP) or the Investments category, depending on whether the annuity has been purchased with registered or unregistered funds. In both of these categories, the gender gap favours men.48 While SBMTs certainly account for some portion of the differential, we cannot determine how much from the data in Table 1, for two key reasons. First, we cannot isolate annuity income from other sources of income in either the Investments or RPP/RRSP category. Second, it is not possible to distinguish between RPP and RRSP income. As we will see in Part V, this distinction matters because the legal rules in Canada governing the use of SBMTs differ with the income source used to purchase an annuity: those purchased with money which has been saved

47. These features include universal portability provisions, and the “child care

drop out” provisions which permit parents to exclude certain years in which they had young children from benefit calculations. See Monica Townson, Independent Means: A Canadian Woman’s Guide to Pensions and Secure Financial Future (Toronto: MacMillan Canada, 1997) at 73–84. Table 1 indicates that in 2009, women collected CPP/QPP pensions that were 79.29% of men’s. 48. In 2009, the gap in the investment category stood at about 27%, up from

22% in 2005. In the RPP/RRSP category, the comparable figure is about 34%, down from 39.5%.

E Shilton 401

within RPPs cannot be priced using SBMTs, while those purchased with RRSP money can be.49

We can nevertheless generate a reasonable estimate of the impact of SBMTs on women’s retirement income in Canada by correlating the insurance industry’s information on annuity premiums and payments with government data. The Canadian Life and Health Insurance Association reports that in 2009, the insurance industry paid out $26.4B in annuity payments.50 Industry data does not tell us how much of this money went to women. It is probable, however, that as much as half of it did—about $13.2B.51 While industry data likewise does not tell us how much of that total of $13.2 billion came from annuities which had been priced on the basis of SBMTs, it is reasonable to estimate that about 70 per cent of it did—some $9.24B.52 If we discount

49. An additional factor is the date when the funds were accumulated; some

annuities now in pay were purchased with pension funds accumulated prior to 1987—that is, before prohibitions on SBMTs within pension plans became widespread. 50. “Key Statistics 2010”, online: Canadian Life and Health Insurance

Association <http://www.clhia.ca/domino/html/clhia/CLHIA_LP4W_LND_Webstation.nsf/resources/The+Industry/$file/KeyStats2010_EN.pdf>. 51. While individual women collect less than 50% of male income in both the

investment and registered plan categories, there are significantly more women than men aged 65 and over in Canada. See “Population Projections for Canada, Provinces and Territories: 2009 to 2036”, online: Statistics Canada <http://www.statcan.gc.ca>. 52. This estimate, necessarily a rough one, is based on continuing data

differentiating between group and individual premiums with data on how much overall premium money had its source in RRSPs. The estimate assumes that annuities were priced based on SBMTs wherever legally possible. See supra note 6. While CLHIA data does not directly break down the total $26.4B into group and individual annuity payments, it does tell us that new annuity premiums for 2009 were divided approximately 47/53 between individual and group annuities, and that RRSPs accounted for approximately 41% of the total premiums in both categories: “Key Statistics 2010”, supra note 50. My estimate assumes that the same breakdown can be applied to annuities in pay. Most, if not all, individual annuities are sex-priced (47%). Most group annuities are purchased from pension funds, but the CLHIA estimates that about 8% of those purchases were made with funds accumulated in jurisdictions that permit sex-based pricing, and that “a substantial portion” of the remainder were purchased with pension funds which are vulnerable to sex-based pricing because they were accumulated prior to 1987.

402 (2012) 37:2 Queen’s LJ

that amount by 12.5%, which is roughly the average sex-based price differential for annuities in Canada,53 we get the figure of $1.155 billion a year, which would appear to be what sex-based annuity pricing costs Canadian women each year. It is likely that the most harmful impact of SBMTs falls on women without male partners (those who are single, divorced, separated or widowed, and those in same-sex relationships) because women in heterosexual relationships may opt for a “custom” version of unisex rate pooling by purchasing “joint and survivor” annuities with their male partners.54

The $1.155 billion associated with SBMTs probably accounts for only about 3.7% of the overall gender income gap for Canadians age 65 and older.55 The other 96.3 per cent is almost certainly attributable to the fact that women bring into retirement fewer financial resources; they earn proportionately less,56 so they save less, both inside and outside their RRSPs, and consequently they own a disproportionately small share of the country’s wealth.57 Women’s unequal role in labour markets and

Information not published by CLHIA was very helpfully provided by CLHIA staff in email correspondence dated May 17, 2011. See supra note 6. 53. See generally supra note 5. 54. I am aware of no studies on this issue, in Canada or elsewhere. 55. Statistics Canada reports that there were 2 616 400 Canadian women age 65

and older in 2009. “Population Projections”, supra note 51 at 135–38. Based on average 2009 figures for this age group ($37 000 for males and $25 700 for females; see “Income: Social Indicators”, supra note 43), the annual value of the total gender gap is about $31.4 billion. 56. See “Average Earnings by Sex and Work Pattern” (15 June 2011), online:

Statistics Canada <http://www.statcan.gc.ca> (summary table showing women earning 64.5 cents to the male dollar in 2008). The sharp jump to 68.8 cents in 2009 is largely accounted for by a significant drop in male earnings in 2009. 57. Margaret Denton & Linda Boos, Gender Inequality in the Wealth of Older

Canadians, QSEP Research Report No 413 (Hamilton: McMaster University Research Institute for Quantitative Studies in Economics and Population, 2007). These researchers found that on average, the net worth of Canadian women age 45 and over is 64% of men’s. Their median net worth is even lower—some 58% of men’s.

E Shilton 403

in families during their pre-retirement years exacts a high and continuing price in retirement.58

While the contribution of SBMTs to the pension gender gap is a modest one, it should not be ignored. Canada’s annuity market is relatively small compared to the amount of capital that has accumulated in pension plans and pension savings instruments in general,59 but it is growing,60 and that growth is likely to be in income categories in which it is currently considered lawful to price annuities on the basis of SBMTs. As enrolment in employment pension plans has dropped off, particularly in the private sector, RRSP contributions have been climbing steadily.61 We can only speculate on how much of these contributions will end up invested in annuities. It is predictable, however, that much more money will find its way into annuities than we have seen in the past, both from RRSPs62 and from other private savings, as

58. Ellie D Berger & Margaret A Denton, “The Interplay between Women’s

Life Course Work Patterns and Financial Planning for Later Life” (2004) 23:5 Canadian Journal on Aging S81; Denton & Boos, supra note 57; Lynn McDonald, “Gendered Retirement: The Welfare of Women and the ‘New’ Retirement” in Leroy O Stone, ed, New Frontiers of Research on Retirement (Ottawa: Statistics Canada, 2006) 137. For comparable US findings, see Tay K McNamara, Regina O’Grady-LeShane & John B Williamson, The Role of Marital History, Early Retirement Benefits, and the Economic Status of Women (Chestnut Hill, MA: Center for Retirement Research at Boston College, 2003); Olivia Mitchell, Philip Levine & John Phillips, The Impact of Pay Inequality, Occupational Segregation and Lifetime Work Experience on the Retirement Income of Women and Minorities, Report No 9910 (Washington: AARP, 1999). 59. James & Vittas, supra note 5 at 1. 60. “Key Statistics 2010”, supra note 50. 61. Men contribute more to RRSPs than women do, in line with their higher

incomes and overall greater wealth. Statistics Canada reports that in 2008, while 47% of RRSP contributors were women, they made only 39% of total contributions. Vince Ferrao, “Paid Work” in Statistics Canada, supra note 42 at 28. 62. Many employers now provide group RRSPs in lieu of pension plans. See

Brenda Lipsett & Mark Reesor, Employer-Sponsored Pension Plans: Who Benefits?, Working Paper (Quebec: Human Resources Development, 1997). Since group RRSPs are governed by the same statutory rules as individual RRSPs, annuities purchased out of group RRSP accumulations will also likely be priced using SBMTs. Whether groups RRSPs will become less attractive with the introduction of the newly-created Pooled Retirement Pension Plans is an open question.

404 (2012) 37:2 Queen’s LJ

pensionless retirees seek to replicate the security of employment pension plans. If legal rules do not change, the direct negative impact of SBMTs on women’s retirement incomes is likely to increase substantially over time, and retirement income inequality along with it.63

Against this historical and factual background, I now turn to a detailed examination of the legal rules governing the use of SBMTs, beginning with how US and European courts have treated those rules in the context of employment pension plans.

III. Legal Rules, SBMTs and Employment Pension Equality: The US and European Case Law

A. “Fair Discrimination” Confronts Individual Rights: The US Cases

As we have seen, the problem of SBMTs generated an energetic academic debate in the 1970s and 1980s, with insurance and law and economics scholars who championed “fair discrimination” facing off against civil rights-oriented scholars

Likewise open is the question of whether insurers will be permitted to use SBMTs to price annuities purchased from funds accumulated inside PRPPs. Enabling legislation—Bill C-25, An Act relating to pooled registered pension plans and making related amendments to other Acts, 1st Sess, 41st Parl, 2011 (first reading 17 November 2011)—is now before Parliament. However, that Bill does not answer the question, although regulations are soon expected to do so. Provinces are expected to adopt companion legislation to Bill C-25 over the next few years. 63. In addition to the direct impact of SBMTs on Canadian women, we need to

be aware of their indirect impact. As we will see in Parts IV and V, under existing law, SBMTs may continue to play a role within employment pension plans for funding purposes. There are no studies evaluating this indirect impact, but it is likely to be negative for female-dominated workplaces, where SBMTs make overall pension costs higher. It is predictable that higher costs will operate as a disincentive to employers to offer plans in female-dominated workplaces, or alternatively, as an incentive to reduce the value of the benefits provided in these workplaces for male and female employees alike. See Bernstein & Williams, supra note 37; Benston, “The Economics of Gender Discrimination”, supra note 19 at 532–36.

E Shilton 405

who promoted individual rights.64 This academic debate was largely fostered by a wave of litigation both in the US and in Europe.65 The litigation did not directly target insurance companies; instead, it targeted employment pension plans in which assumptions that women would live longer meant that they either contributed more or received lower benefits. These cases forced the courts to explore the meaning of equal pension entitlements in the context of statutory equality guarantees in the US and quasi-constitutional equality guarantees in Europe.

Legal challenges arose first in the US, based on Title VII of the federal Civil Rights Act, 1964. Title VII generally prohibits employment discrimination based on sex, but expressly excepts differences in compensation that comply with the older Equal Pay Act, which permits differences in pay between male and female employees if they are based on “factors other than sex”. Initially, guidelines used by administrators who enforced equal pay laws took an either-or approach to sex-based pension practices; plans were deemed to meet “equal pay” standards if they imposed equal employee contributions or paid equal benefits, but were not required to do both. When the Equal Employment Opportunities Commission took over the administration of equal pay laws in 1972, it introduced new guidelines that required both equal contributions and equal benefits.66 The litigation which followed ultimately culminated in two decisions, Manhart67 and Norris,68 in which the US Supreme Court banned employer use of both sex-

64. Benston provides exhaustive citations to the pre-Manhart academic

discussion. “The Economics of Gender Discrimination”, supra note 19 at n 10. See also Wortham, “Insurance Classification”, supra note 21 at n 31 (providing update to these citations). 65. The intense academic interest in the issue may be attributable in part to the

fact that much of this litigation was directed against TIAA-CREF, the pension plan established by the Carnegie Foundation for academics. See Judith E Tytel, “TIAA-CREF and the Sex-Based Mortality Table Controversy” (1980) 7 JC & UL 119; Key, supra note 26. TIAA-CREF was not involved in the cases which subsequently reached the US Supreme Court. 66. Tytel, supra note 65 at 120–21; Bernstein & Williams, supra note 37 at 1208–

10. 67. Manhart, supra note 26. 68. Arizona Governing Committee for Tax Deferred Annuity and Deferred

Compensation Plans v Nathalie Norris, 463 US 1073 (1983).

406 (2012) 37:2 Queen’s LJ

based employee contribution rates and sex-differentiated pension benefits within employment pension plans.

Manhart was the first case to challenge the practice of requiring female employees to make higher pension contributions than men in order to receive equal periodic benefits.69 The plan at issue was a self-administered contributory defined benefit plan with a benefit formula based on salary and length of service. Male and female employees who retired at the same age with the same salary and length of service received the same monthly pension benefit. To earn that benefit, however, female employees had to contribute almost 15 per cent more of their annual salary than male employees.70

In defence of this direct distinction between men and women, the employer made two arguments. First, it asserted that the distinction was based on longevity and not on sex, and therefore complied with the Equal Pay Act. This argument appealed to the two dissenting judges, who were persuaded that sex was simply “the umbrella-constant under which all of the elements leading to differences in longevity are grouped and assimilated”.71 The majority speedily dispatched the argument, adopting the trial judge’s observation that “an actuarial distinction based entirely on sex” cannot be “based on any other factor other than sex”.72

The second string to the employer’s bow was the “equal actuarial value” argument cast in terms very similar to those explored in Part I above. The majority conceded the validity of the argument as it applied to women as a group, but they saw

69. Requiring higher contributions from women in return for equal benefits was

an unusual approach. Much more common was the practice of making equal employer contributions but paying women lower benefits. See Key, supra note 26 at 1–2. 70. The cumulative impact of this distinction was considerable; evidence before

the court showed that in one specific case, a woman employee had to contribute almost 50% more than a similar male employee ($18 171.40 compared to $12,843.53) in order to buy the same periodic pension. Manhart, supra note 26 at n 5. 71. Ibid at 727. Burger CJ and Rehnquist J, dissenting, found no discrimination,

characterizing the impact of the majority decision as “revolutionary and discriminatory” against men. Ibid at 726. 72. Ibid at 712–13.

E Shilton 407

Title VII as creating rights for individuals rather than groups. Under the Manhart plan, every woman paid a higher premium for her pension than every man at her pay level; only some women actually benefitted by collecting their pensions longer than men.73 In the court’s view of Title VII, individual women were not to be penalized by the ascription of group characteristics: “[e]ven a true generalization about the class is an insufficient reason for disqualifying an individual to whom the generalization does not apply”.74 Title VII requires that “thoughtful scrutiny of individuals” replace “traditional assumptions” about classes of individuals—no matter how valid, efficient or useful those assumptions may be—when making employment decisions that affect protected categories: “[t]he basic policy of the statute requires that we focus on fairness to individuals rather than fairness to classes”.75

The Court recognized that in the insurance context, insistence on an individualized approach might produce group “unfairness”. Indeed, it saw the potential for conflict between its individualistic focus in Manhart and its earlier recognition in Willie S. Griggs et al., Petitioners, v. Duke Power Co.76 that disparate impact on a protected group may be unlawfully discriminatory.77 The Court

73. Ibid (“there is no assurance that any individual woman working for the

Department will actually fit the generalization on which the Department's policy is based. Many of those individuals will not live as long as the average man” at 708). 74. Ibid. 75. Ibid at 709. 76. 401 US 424 (1971). 77. In Brilmayer et al, “Sex Discrimination”, supra note 25 at 508, the authors

reconcile the individualistic approach in Manhart with the group-oriented disparate impact approach espoused in Griggs by arguing that where claims of disparate treatment (what Canadian law would normally call “direct discrimination”) and claims of disparate impact (what Canadian law would normally call “indirect discrimination”) collide, disparate treatment claims must always prevail. Indeed, they go so far as to argue that with respect to equal pay, only disparate treatment (i.e. direct discrimination) is prohibited. Ibid at 516, 521–22. In the wake of 1991 amendments to Title VII expressly codifying disparate impact case law such as Griggs, the argument that a disparate impact defence can never be used to justify disparate treatment was rejected by the US Supreme Court in Ricci v DeStefano, 557 US (07-1428) (2009). The Court held

408 (2012) 37:2 Queen’s LJ

was content to live with this potential contradiction, taking comfort in the fact that insurance is fundamentally a group concept which always involves “subsidization”. Anti-discrimination law does not preclude insurers from creating risk groups, but it does ask where group boundaries can legitimately be drawn. The court emphasized that the power to answer this question belongs to Congress, and not to the market:

Congress has decided that classifications based on sex, like those based on national origin or race, are unlawful. Actuarial studies could unquestionably identify differences in life expectancy based on race or national origin, as well as sex. But a statute that was designed to make race irrelevant in the employment market . . . could not reasonably be construed to permit a take-home-pay differential based on a racial classification.78

By the same token, differences in compensation based directly on sex are unlawful as a matter of statutory policy, regardless of their actuarial validity.

Manhart banned sex-based differences in employee contributions, but explicitly did not address the correlative question of whether Title VII required employers to equalize periodic pension benefits. This was the core issue in Norris, which reached the US Supreme Court five years later. Norris involved a deferred compensation plan (the functional equivalent of a defined contribution plan)79 in which employees could elect to set aside a certain percentage of their wages until retirement. Since each employee chose what percentage of wages to defer, there was no issue of direct sex discrimination in their contributions. However, employees who chose to annuitize their deferred compensation accounts on retirement were required to purchase annuities from insurers selected by the employer. All of these insurers used SBMTs, so women employees retiring at the same

that disparate treatment may be justified where the defendant can show a “strong basis in evidence” that it is necessary to prevent disparate impact. 78. Manhart, supra note 26 at 709. 79. Plans that do not require an employer contribution could not be registered

as a pension plan in Canada. This was the type of plan involved in Norris, supra note 68.

E Shilton 409

age as male employees got lower periodic benefit payments for equal amounts of deferred compensation.

Once again, the court’s analysis focused on the individual woman. As in Manhart, the Norris employer made the equal actuarial value argument, but the court emphasized its circularity:

[P]etitioners incorrectly assume that Title VII permits an employer to classify employees on the basis of sex in predicting their longevity. Otherwise there would be no basis for postulating that a woman’s annuity policy has the same present actuarial value as the policy of a similarly situated man even though her policy provides lower monthly benefits.80

The court saw no meaningful legal distinction between Manhart’s sex-based contributions, and Norris’s sex-based benefits:

If a woman participating in the Arizona plan wishes to obtain monthly benefits equal to those obtained by a man, she must make greater monthly contributions than he, just as the female employees in Manhart had to make greater contributions to obtain equal benefits.81

The equal actuarial value defence could not prevail against Title VII’s policy objective—to make sex irrelevant in employment decisions.

The decisions in Manhart and Norris largely did away with the use of sex-based mortality assumptions in employment pensions in the US. The practices of the insurance industry outside employment pension plans, however, were not affected. Before turning to the issue of insurance practices, it is useful to explore how the European Court of Justice has dealt with the use of SBMTs in employment pension plans, as its decisions on that matter are the direct precursor to its 2011 Test-Achats decision banning industry use of SBMTs in Europe.

80. Ibid at 1083. 81. Ibid at 1082.

410 (2012) 37:2 Queen’s LJ

B. Sex-Based Pension Distinctions and “Equal Pay”: The European Court of Justice

The issue of SBMTs reached the ECJ in the 1990s, a decade after the US Supreme Court decided Norris. As in the US, the issue emerged as a question of “equal pay” within employment pension plans. However, in Europe, the equal pay guarantee has quasi-constitutional status under Article 119 of the Treaty of Europe, rather than simply statutory status. Article 119 is also supplemented by Directive 86/387, which generally mandates equal treatment in pension benefits.82 During the 1980s, the ECJ made several key decisions confirming that Article 119 applied to a wide range of employment-related pension plans, which effectively prohibited employers from providing pension benefits to male employees that were not provided equally to female employees.83

In 1993, two cases, Coloroll Pension Trustees84 and Neath v. Steeper,85 brought the SBMT issue directly before the ECJ. Unlike

82. See EEC, Council Directive 86/378/EEC of 24 July 1986 on the

Implementation of the Principle of Equal Treatment for Men and Women in Occupational Social Security Schemes, [1986] OJ, L 225/40, arts 6(1)(h)–(i), 9(c). The Directive contains certain express exceptions, some of which relate to the use of actuarial calculation factors in defined contribution plans, and which also expressly permits the use of sex-differentiated employer contributions if their purpose is to equalize benefits. The content of the Directive is discussed in Ten Oever v Stichting Bedrijfspensioenfonds voor het Galenwassers-en Schoonmaakbedrijf, C-109/91 (1993), Advisory Opinion of Advocate General Van Gerven, [1993] ECR I-4893 at para 29 [Van Gerven]. 83. Worringham and Humphreys v Lloyds Bank, C-69/80, [1981] ECR I-768;

Bilka-Kaufhaus GmbH v Karin Weber von Hartz, C-170/84, [1986] ECR I-1620; Barber v Guardian Royal Exchange Assurance Group, C-262/88, [1990] ECR I-1944. Much of this history is discussed in Sandra Fredman, “The Poverty of Equality: Pensions and the ECJ” (1996) 25:2 Indus LJ 91. See also Van Gerven, supra note 82; Evelyn Ellis & Philip Morell, “Sex Discrimination in Pension Schemes: Has Community Law Changed the Rules?” (1982) 11:1 Indus LJ 16. 84. Coloroll Pension Trustees v Russell and Others, C-200/91, [1994] ECR I-04389

at para 9. 85. Neath v Steeper, C-152/91, [1993] ECR I-6935 (both this case and Coloroll,

ibid, were referred from the UK). See also Ten Oever v Stichting Bedrijfspensioenfonds voor het Galenwassers-en Schoonmaakbedrijf, C-109/91, [1993] ECR I-04879 at paras 3–5.

E Shilton 411

the US cases, neither Coloroll nor Neath involved sex-differentiated contributions or periodic benefits—both of the plans at issue were defined benefit plans that offered equal benefits to male and female employees in return for equal employee contributions. Instead, the focus of Coloroll and Neath was on the use of SBMTs in calculating the “transfer value” or the “commuted value” of the plan. The claims came not from female employees but from male employees.

In order to understand these claims, it is necessary to understand the terminology used in the cases. “Transfer values” are values calculated when a pension plan member leaves the plan, but is not immediately retiring on pension and wishes to transfer his or her pension credits to another plan. “Commuted values” involve similar calculations relating to the payments made directly to an employee who is leaving the plan, but where the employee seeks a buy-out of his or her credits rather than a transfer. Both transfer value and commuted value represent the capital value of the employee’s pension entitlement: i.e. the employee’s entitlement to the income stream or the defined benefit. This valuation takes into account life expectancy. If SBMTs are used in calculating longevity, the pension entitlement of a male employee will be less than that of a female employee, since the mortality tables reflect a shorter lifespan for males. The Coloroll and Neath claimants therefore argued that the use of SBMTs in calculating the capital value of their pension entitlements violated their right to equal pay.

The Advocate General’s opinion in Coloroll and Neath essentially followed the reasoning in Manhart and Norris. The Advocate General saw both commuted values and transfer values as pension benefits, and concluded that Article 119 did not permit the group-based assumptions inherent in SBMTs to apply to pension benefits.86 The ECJ took its own path, however, holding that Article 119 applied neither to commuted values nor to transfer values. The Court reasoned that the contributions required of employees are “an element of their pay, since they are

86. Van Gerven, supra note 82. Under ECJ procedure, advocates general provide

legal opinions to the court. Although these opinions are non-binding, they are normally followed.

412 (2012) 37:2 Queen’s LJ

deducted directly from an employee’s salary, which by definition is pay”.87 Accordingly, employee contributions could not be differentiated on the basis of sex. However, the court saw employer contributions as conceptually quite different. In defined benefit plans, employer contributions are designed to “ensure the adequacy of the funds necessary to cover the cost of the pensions promised”.88 “It follows”, observed the court, “that, unlike periodic payment of pensions, inequality of employers’ contributions paid under funded defined benefit schemes, which is due to the use of actuarial factors differing according to sex, is not struck at by Article 119”.89 As the court saw it, these unequal contributions are not “pay”, and neither is the “conversion of part of the periodic pension into a capital sum and the transfer of pension rights, the value of which can be determined only by reference to the funding arrangements chosen”.90

The reasoning behind this conclusion is troubling. If pension benefits themselves are “pay”, as they clearly are, the argument that the commuted value of such benefits is also “pay” deserves more careful consideration than the court gave to it. Nevertheless, in a world in which SBMTs still held sway both in financial markets and for plan funding purposes, the decision had the practical merit of matching capital values to the real costs of replicating periodic benefits offered by the plans. If the plaintiffs had succeeded in obtaining the same lump sum as female members of the plans with the same pay and work history, they could have taken those lump sums to market and converted them into larger SBMT-priced annuities than their female counterparts—an outcome that would have been clearly inconsistent with the court’s equal pay jurisprudence.

Coloroll and Neath left the European jurisprudence in much the same state as the American jurisprudence, prohibiting the use

87. Neath, supra note 84 at para 31. 88. Ibid. 89. Ibid at para 32. 90. Ibid at para 33. See also Coloroll, supra note 84 at paras 79–85. This

conclusion is, of course, the diametric opposite of Kimball’s conclusion that employer contributions are the only pension element that can properly be characterized as “pay”. See discussion in Part II below.

E Shilton 413

of SBMTs by employers within employment pension plans to the extent that they resulted in sex-based distinctions in contributions and benefits. In neither jurisdiction, however, did the jurisprudence directly affect insurance practices. The majority in Manhart emphasized that its holding did not apply to the sale of annuities on open financial markets,91 and the court in Norris observed that its “judgment will in no way preclude any insurance company from offering annuity benefits that are calculated on the basis of sex-segregated actuarial tables”.92 These obiter comments did not endorse existing insurance practices; they simply reflected the fact that the insurance industry was not subject to federal anti-discrimination laws.93 Similarly in Europe, Neath and Coloroll did not approve of the use of SBMTs in plan funding, but simply held that Article 119 did not apply to the calculation of either transfer values or commuted values. The jurisprudence nevertheless gave insurers an effective immunity for the use of SBMTs outside the employment pension context—an immunity they continued to use widely.

In 2011, the ECJ decision in Test-Achats stripped insurers of that immunity in Europe. I turn now to that decision.

91. In addition, because the plan in Manhart was self-administered, the decision

did not “call into question the insurance industry practice of considering the composition of an employer's work force in determining the probable cost of a retirement or death benefit plan”. Manhart, supra note 26 at 718. 92. Norris, supra note 68 at 1087. As part of its defence in Norris, the employer

had attempted to pass off responsibility to the insurers who sold the discriminatory annuities. The court refused to let the employer off the hook. It held that Title VII required that employment benefits be non-discriminatory. Since the employer selected and contracted for the plan it offered to its employees, it was liable for the discrimination imbedded in that plan. 93. US insurers are regulated by a plethora of state insurance laws, most of

which do not explicitly ban sex discrimination in the setting of premiums or the payment of benefits. See Wortham, “Economics”, supra note 16 at 850. At the time Norris was argued, there was a bill before Congress that would have banned all sex-based insurance practices. This bill was not passed. See Wortham, “Insurance Classification”, supra note 21 at 364–65; Jill Gaulding, “Race, Sex, and Genetic Discrimination in Insurance: What's Fair?” (1995) 80:6 Cornell L Rev 1646 at n 49.

414 (2012) 37:2 Queen’s LJ

IV. Taking on the Insurance Industry: Association belge des Consommateurs Test-Achats ASBL94

While the debate about SBMTs subsided in the US courts once the issue was resolved for employment pension plans, the controversy over sex-based insurance practices continued in Europe. In 2004, the Council of the European Union weighed into that controversy with Directive 2004/113, which was designed to combat sex discrimination in the provision of goods and services.95 Insurance was a key target of the Directive, which generally contemplated that sex would be phased out as a factor in determining premiums and benefits after a transitional period designed “to avoid a sudden readjustment of the market”. The Directive was rendered ambiguous, however, by article 5(2), an open-ended “derogatory” provision allowing member states to continue to permit sex-based differentials in both insurance premiums and benefits “where the use of sex is a determining factor in the assessment of risk based on relevant and accurate actuarial and statistical data”.96 Numerous member states, including Belgium, took advantage of this opt-out provision to pass legislation permitting the use of SBMTs in life insurance and annuity products. Opponents of Belgium’s law launched a challenge before its national Constitutional Court, arguing that the Directive was inconsistent with article 6 of the Treaty on European Union and articles 21 and 23 of the European Charter, which together guarantee respect for fundamental rights, prohibit

94. Test-Achats, supra note 7. The Association belge des Consommateurs Test-

Achats is a consumer advocacy organization. 95. The use of SBMTs was one of the principal targets of the Directive. See

Eugenia Caracciolo Di Torella, “The Goods and Services Directive: Limitations and Opportunities” (2005) 13:3 Fem Legal Stud 337 at 343. See also Thiery & Van Schoubroeck, supra note 30 at 190–91. 96. EC, Council Directive 2004/113/EC of 13 December 2004 Implementing the

Principle of Equal Treatment Between Men and Women in the Access to and Supply of Goods and Services, [2004] OJ, L 373/37, art 5(2). There were a variety of qualifications to the right of derogation, including the requirement that the statistics relied on must be relevant, accurate, up to date and publicly available.

E Shilton 415

discrimination on the basis of sex and mandate equality between men and women.97 In 2009, the challenge was referred to the ECJ.98 In September of 2010, Advocate General Kokott issued her opinion that article 5(2) of the Directive violated the principle of equal treatment and should be invalidated.99 In May of 2011, the ECJ came to the same conclusion.100

Test-Achats required the ECJ to focus squarely, for the first time, on whether the sex-based practices of the insurance industry met European constitutional equality guarantees. In addressing this question, the ECJ did not reject sex as an acceptable classifier per se. Instead, consistent with its earlier jurisprudence, it acknowledged that “the principle of equal treatment requires that comparable situations must not be treated differently and different situations must not be treated the same way, unless such treatment is objectively justified”.101 The key question posed by article 5(2) of the Directive, then, was whether an insurer’s use of statistical measures of longevity provided the “objective justification” required by the principle of equal treatment. In answering that question, the advocate general had largely adopted the individual rights perspective laid out in the US jurisprudence. She recognized that in predicting longevity, insurers must carry out risk assessment on a group rather than on an individual basis,102 but she found sex to be an “inappropriate” ground for drawing group boundaries, at least in the absence of “any clear biological differences between insured persons”.103 The ECJ

97. Under European Law, directives are subordinate legislation which must

conform to European constitutional law. 98. For an explanation of the relationship between the ECJ and national courts

in Europe in the context of equal pay litigation, see Karen J Alter & Jeannette Vargas, “Explaining Variation in the Use of European Litigation Strategies: European Community Law and British Gender Equality Policy” (2000) 33:4 Comparative Political Studies 452. 99. Association belge des Consommateurs Test-Achats ASBL and Others (2010),

Opinion of Advocate General Kokott, C-236/09 [Opinion, Test-Achats]. 100. Test-Achats, supra note 7. 101. Ibid at para 28. For a general analysis of the ECJ’s approach to equality, see

Catherine Barnard & Bob Hepple, “Substantive Equality” (2000) 59:3 Cambridge LJ 562. See also Fredman, supra note 83. 102. Opinion, Test-Achats, supra note 99. 103. Ibid at paras 50-52.

416 (2012) 37:2 Queen’s LJ

followed a less complex path to the same practical conclusion. The Directive’s premise, in its view, was that men and women were comparable for the purposes of insurance premiums and benefits. article 5(2), then, was inconsistent with this premise, and potentially undermined the principle of equal treatment reflected in the Directive.104 The ECJ was not prepared to tolerate the temporally open-ended exemption reflected in article 5(2), but it was prepared to accept the need for transitional measures to protect the stability of insurance markets. Accordingly, the ECJ declared article 5(2) invalid, effective December 2012.105

To summarize, US law prohibits the use of SBMTs in employment pension plans, at least to the extent that they produce differential employee contribution or benefit levels.106 To date, however, there have been no federal restrictions placed on their use by insurers in financial markets. In Europe, the status of SBMTs within employment-related pension plans continues to be governed directly by article 119 and Directive 86/387, which combine to outlaw SBMTs that produce sex-differentiated employee contributions and benefits, but leave loopholes for SBMTs to continue affecting transfer/commuted values and (possibly) other funding issues. Once Test-Achats is fully absorbed, however, SBMTs may not be used to affect insurance premiums and benefits in open markets.107

Against this backdrop, I now turn to the status of SBMTs under Canadian law.

104. Test-Achats, supra note 7 at paras 30–31. 105. Ibid at para 34. 106. With the exception of the Supreme Court’s obiter comment on the costing

of pension plans (see Norris, supra note 68 at 1087), US case law does not address the question of the extent to which SMBTs may affect pension plan funding. 107. Whether Test-Achats will also limit insurance practices relating to pension

plan funding remains to be determined. Some major UK and international law firms have suggested that it may. See e.g. “The Gender Directive and the ECJ Decision on Gender Specific Insurance Premiums and Benefits”, online: Pitmans LLP <http://www.pitmans.com>; “ECJ Rules for Equal Treatment in Insurance Case”, online: Linklaters LLP <http://www.linklaters.com>.

E Shilton 417

V. SBMTs in Canada

A. The Statutory Framework

In Canada, the use of SBMTs is governed by a complex set of statutory provisions that largely mirror the US rules,108 including the distinction between rules that apply to employers and to insurers.109 For much of this discussion, I focus on the law of Ontario, which is reasonably typical of Canadian regulation in this area.110

The starting point for the analysis is the Ontario Human Rights Code.111 Part I of the Code establishes a set of positive anti-discrimination rights, two of which are directly applicable to insurance contracts: a right to equal treatment “with respect to

108. Canada’s legal rules were cemented in place in the 1980s, after the decisions

of the US Supreme Court in Manhart, supra note 26 and Norris, supra note 68. It is logical to assume that the similarity between the two sets of rules is deliberate, in view of the continental nature of the marketplace in financial services. 109. In Canada, employment relationships are regulated primarily at the

provincial level, with federal law applying only to employment within federal jurisdiction. Insurance is a split jurisdiction, with the federal government’s responsibilities largely confined to prudential regulation. Issues of risk classification are governed by provincial legislation. Federal law governs income taxation, which has an important influence on the design of retirement income instruments but is not discussed in this paper. 110. Regulatory provisions do differ from province to province, and close

analysis would be required to assess the impact of these differing legal frameworks on individual pensioners. Regulations under the Supplemental Pension Plans Act, RSQ, c R-15.1 require the use of SBMTs within employment pension plans for the purposes of certain funding calculations, including commuted values. See Regulation Respecting Supplemental Pension Plans, RRQ, c R-15.1, r 1, s 67.4. Quebec’s Charter of Human Rights and Freedoms, RSQ, c C-12 excludes from its anti-discrimination prohibitions distinctions made on the basis of sex “where the use thereof is warranted and the basis therefore is a risk determination factor based on actuarial data” at s 20.1. Legal rules in Alberta, Newfoundland and Prince Edward Island may also permit sex-based distinctions within pension plans that would be prohibited by Ontario law. See Q & A Prepared by the Canadian Institute of Actuaries for CAPSA Regarding Commuted Value of Pensions (Toronto: Canadian Institute of Actuaries, 2005) at 4. 111. RSO 1990, c H 19.

418 (2012) 37:2 Queen’s LJ

services, goods and facilities”112 without regard to sex, and a right to “contract on equal terms without discrimination”.113 These are prima facie rights, qualified in Part II by a series of exceptions and defences, including this specific “insurance” defence established in section 22:

The right under sections 1 and 3 to equal treatment with respect to services and to contract on equal terms, without discrimination because of age, sex, marital status, family status or disability, is not infringed where a contract of automobile, life, accident or sickness or disability insurance or a contract of group insurance between an insurer and an association or person other than an employer, or a life annuity, differentiates or makes a distinction, exclusion or preference on reasonable and bona fide grounds because of age, sex, marital status, family status or disability.114

In other words, section 22 does not give insurers a defence to complaints based on most of the types of discrimination prohibited by sections 1 and 3 of the Code (race, ancestry, place of origin, colour, ethnic origin, citizenship, creed or sexual orientation), but it does allow an insurer to defeat a complaint of sex discrimination (or discrimination for age, marital or family status, or disability) if it proves that a distinction drawn on the basis of one of these types of discrimination was drawn “on reasonable and bona fide grounds”.

For insurers in Ontario, the regulatory picture is completed by the provincial Insurance Act. Although it is an offence under that Act to “discriminate unfairly” on the basis of race and religion,115 it contains no specific prohibition on the use of sex-based distinctions. Regulations governing fair practices require only that insurers refrain from “unfair discrimination between individuals of the same class and of the same expectation of life”.116 This language appears to leave insurers free to establish

112. Ibid, s 1. 113. Ibid, s 3. 114. Ibid, s 22. 115. Unfair or Deceptive Acts or Practices, supra note 27, s 140. It is interesting to

note that this not a per se prohibition. It implicitly recognizes the possibility of “fair” discrimination on the basis of race and religion. 116. Ibid, s 1(2).

E Shilton 419

their own classes and make their own determinations about life expectancy.117

Group insurance contracts made by employers with insurers in order to provide coverage for employees cannot shelter under the insurer’s defence to sex discrimination set out in section 22 of the Human Rights Code, but are governed by a different set of prima facie rights and defences, under the Code. Section 5 protects “a right to equal treatment with respect to employment without discrimination” on the basis of sex, and the corresponding defence is found in section 25(2):

The right under section 5 to equal treatment with respect to employment without discrimination because of sex, marital status or family status is not infringed by an employee superannuation or pension plan or fund or a contract of group insurance between an insurer and an employer that complies with the Employment Standards Act, 2000 and the regulations thereunder.118

The Employment Standards Act, 2000 (ESA), like the Human Rights Code, establishes a prima facie obligation on employers to ensure that their benefit plans (including pension plans) are free of discrimination on the basis of sex.119 This obligation is qualified by a provision in the regulations which expressly allows employers to make sex-differentiated contributions to pension plans “if the differentiation is made on an actuarial basis because of an employee’s sex and in order to provide equal benefits under the plan”.120 The regulation therefore highlights the primary obligation on employers to provide equal benefits at equal cost to the employee. By implication, however, it also acknowledges a role for SBMTs in plan funding. The scheme of the Code and the

117. Ibid. 118. Supra note 111. Section 25 goes on to spell out in sub-section 2.3 that “[f]or

greater certainty, subsections 2 and 2.1 apply whether or not age, sex or marital status in the Employment Standards Act, 2000 or the regulations under it have the same meaning as those terms have in this Act”. 119. Employment Standards Act, 2000, SO 2000, c 41, s 44 (the prima facie right

to protection from discrimination extends not just to employees, but also to their beneficiaries, survivors and dependents). 120. O Reg 286/01, s 2(1) [emphasis added]. Pre-1987 employment is generally

exempt.

420 (2012) 37:2 Queen’s LJ

ESA is reinforced by the Pension Benefits Act (PBA),121 which spells out a number of specific non-discrimination standards, including an explicit prohibition against sex discrimination in benefits and employee contributions. Like the ESA, the PBA contemplates the continued use of SBMTs to “provide for employer contributions that vary according to the sex of the employee”, but only for the limited purpose of producing non-discriminatory benefit outcomes.122

Because the Ontario law gives insurers more scope for discrimination than it gives to employers, various anomalies arise at the intersection points between employment pension plans and insurance practices. These anomalies permit insurers to impose higher costs on a plan because of the gender of plan members as long as the purpose is to provide equal pensions for women, and as long as the consequences of those higher costs are distributed on a gender-neutral basis. For example, the annual premium charged to employers for an insured pension plan may be higher for a predominantly female workforce.123 Anomalies also arise where an individual employee leaves a pension plan before retirement and transfers accumulated pension credits to an individual locked-in account. Because the source of the funds is a registered pension plan, annuities purchased from these individual accounts are governed by pension rules rather than insurance rules: there must be a gender-neutral outcome. In contrast to European law, Ontario law clearly prohibits pension plans within its jurisdiction from using SBMTs to calculate transfer values/commuted values.124 Because laws governing the use of SBMTs in pension plans vary in other jurisdictions, however, Ontario law provides a regulatory formula for “neutralizing” locked-in funds to produce non-discriminatory outcomes.

121. RSO 1990, c P-8 [PBA]. 122. Ibid, s 52(2). Like O Reg 286/01, supra note 120, s 52(3), this provision

generally applies only to pension rights accumulated from 1987 onward. 123. Employers of such workforces may respond to these higher costs by

declining to provide a pension plan. Alternatively, they may reflect these higher costs in higher employee contribution rates or lower benefits, as long as those higher costs are imposed on all employees. 124. PBA, supra note 121, s 52(1)(b).

E Shilton 421

Regulations under the PBA currently require that where the commuted value of the locked-in funds was originally calculated using SBMTs, SBMTs must be used to price the annuity. If the original amount was calculated using unisex mortality tables, however, unisex tables must be used to price any annuity purchased later.125

B. “Reasonable and Bona Fide” Insurance Practices

Although the rules governing the use of SBMTs within employment pension plans are complex, their bottom line is clear: SBMTs may not be used to produce gender-based differences in employee contributions or employee benefits, including transfer/commuted values, although they may continue to play a role in plan funding. Outside of pension plans, insurers have much more scope to use SBMTs. There is no explicit statutory sanction for the practice. As we will see, its legality depends on whether it is “reasonable and bona fide” within the meaning of section 22 of the Human Rights Code. The leading case interpreting section 22 is the 1992 decision of the Supreme Court of Canada in Zurich Insurance Co. v. Ontario (Human Rights Commission),126 where the claimant challenged the insurance industry practice of charging single males under the age of 25 significantly higher car insurance premiums than single females of comparable age. A Board of Inquiry under the Code found that basing auto insurance premiums on the factors of sex, age and marital status was not “reasonable and bona fide”. The company’s appeal eventually made its way to the Supreme Court of Canada, where a majority, in a decision authored by Sopinka J., quashed the Board’s decision and upheld the premiums.

Rejecting the more onerous “reasonable necessity” test for bona fides applied in the employment context,127 Sopinka J. set

125. RRO 1990, Reg 909, s 22. 126. [1992] 2 SCR 321. 127. Ontario (Human Rights Commission) v Etobicoke (Borough), [1982] 1 SCR

202 (the court held that to satisfy the objective element of the legal test for identifying a bona fide occupational qualification or requirement, an employer would be required to demonstrate that the qualification or requirement was

422 (2012) 37:2 Queen’s LJ

out a unique two-pronged test for exempting distinctions under section 22. First, the distinction must be bona fide in the sense that it “was adopted honestly, in the interests of sound and accepted business practice and not for the purpose of defeating rights protected under the Code”.128 Second, it must be reasonable in the dual sense that “(a) it is based on a sound and accepted insurance practice; and (b) there is no practical alternative”.129 Justice Sopinka defined a “sound” insurance practice as one that the industry has found “desirable to adopt for the purpose of achieving the legitimate business purpose of charging premiums that are commensurate with risk.”130 In his view, establishing automobile insurance premiums on the basis of sex, age and marital status met this test. His decision did reflect some discomfort with the status quo-oriented nature of that conclusion; he cautioned that the “insurance industry must strive to avoid setting premiums based on enumerated grounds”.131 He was nevertheless prepared to defer to industry business practice, despite sharp criticism from dissenting members of the Court, who observed (rightly, in my view) that “[n]o human rights legislation could ever attain its objectives if discrimination could be justified by the self-serving claim that a practice ‘has always been done this way’”.132

Content with the outcome in Zurich, the insurance industry has evidently seen no need to heed the Supreme Court’s admonition to get on with the process of developing “practical alternatives” to distinctions based on prohibited grounds. While some provinces have prohibited sex-based pricing in auto

“reasonably necessary to assure the efficient and economical performance of the job” at para 8). See also Zurich, supra note 126 at para 19. 128. Ibid at para 24. 129. Ibid at para 23. 130. Ibid. 131. Ibid at para 43. 132. Ibid at para 106, L’Heureux-Dubé J, dissenting. Justice McLachlin echoed

that sentiment in her own dissent; while she acknowledged the difficulty facing an insurer attempting to find practical non-discriminatory alternatives to current practices, she noted that “difficulty alone has never been accepted as an excuse for discriminatory conduct contrary to human rights legislation”. Ibid at para 130.

E Shilton 423

insurance,133 there is as yet no sign of change in Ontario, and no Canadian legislature has moved to restrict the use of SBMTs to price annuities on the open market. The key legal question now is whether that practice can continue in the face of evolving equality standards and changing legal norms in other comparable jurisdictions. If the ECJ decision in Test-Achats prompts a challenge to the legality of SBMTs in Canada, is the challenge likely to succeed, or will SBMTs continue to find shelter under the Zurich holding?

C. SBMTs and Contemporary Canadian Legal Equality Standards

If SBMTs had been the target in Zurich, they arguably would not have survived that decision. Important to the company’s defence in Zurich were the arguments that statistics based on the challenged criteria were the only reliable tools then available for constructing automobile insurance risk categories, and that those statistics had been vetted through a centralized and state-supervised process to establish “fair” rates.134 In the context of life insurance, there is no comparable centralized process for data-gathering or rate-setting. Mortality tables may differ from company to company. When applied within pension plans, they are frequently fine-tuned to reflect the specific experience of the particular plan.135 While the industry clings to the claim that SBMTs are a uniquely valuable tool for risk assessment, practical alternatives have clearly been available since the 1990s in the form of unisex mortality tables, which the industry developed in response to the evolution of legal rules prohibiting discrimination within employment pension plans in Canada and elsewhere.136 If

133. Five Canadian provinces now prohibit the use of sex in establishing auto

insurance rates either in whole or in part, including three with public auto insurance systems—British Columbia, Saskatchewan and Manitoba. 134. There is a detailed discussion of this process in Zurich, supra note 126 at

paras 26–31. 135. Key, supra note 26 at 11. 136. See Society of Actuaries UP-94 Task Force, supra note 25 at 826

(acknowledging that the predecessor to the current UP-94 table (the UP-84 table) was a unisex table, and that while actuaries express a strong attachment to what

424 (2012) 37:2 Queen’s LJ

challenged on the matter today, insurers would face even stronger factual hurdles in arguing that there are no practical alternatives to SBMTs, in light of the fact that European insurers are now developing tools to respond to Test-Achats in a globalized insurance marketplace.

In addition to those factual hurdles, SBMTs would today face a human rights jurisprudence that has evolved since 1992. It is not entirely clear whether the Supreme Court of Canada’s current perspective on discrimination and equality rights in the human rights context would benefit the challengers or the insurers.137 Insurance scholars Trudo Lemmens and Yves Thiery argue that if the Supreme Court of Canada were now confronted with a Zurich problem, it would be much less deferential to the insurance industry.138 They point to the Supreme Court’s important 1999 decision in the Meiorin139 case, where the Court collapsed the distinction drawn in earlier cases between direct and indirect discrimination. Lemmens and Thiery see the demise of that distinction as importing a requirement of individual accommodation into all discrimination claims, including those involving direct discrimination.140 They predict that courts and

they view as the superior accuracy of SBMTs, they acknowledge methodologies for adapting them to situations that call for unisex tables). 137. See Karen Schucher, “Human Rights Statutes as a Tool to Eliminate and

Prevent Discrimination: Reflections on the Supreme Court of Canada’s Jurisprudence” (2010) 50 Sup Ct L Rev (2d) 387 at 397–403 (discussing the Supreme Court’s inconsistent adherence to substantive equality in the human rights context). 138. Trudo Lemmens & Yves Thiery, “Insurance and Human Rights: What Can

Europe Learn from Canadian Anti-Discrimination Law?” in Herman Cousy & Caroline Van Shoubroeck, eds, Discrimination in Insurance (Maklu: Academia-Bruylant, 2007), online: SSRN <http://papers.ssrn.com>. See also Ontario Human Rights Commission, Human Rights in Insurance, Discussion Paper (Toronto: Policy and Education Branch, 1999) at 10–11 (the Commission argues that events have overtaken Zurich Insurance with respect to automobile insurance). 139. British Columbia (Public Service Employee Relations Commission) v British Columbia Government Services Employees’ Union, [1999] 3 SCR 3 [Meiorin]. 140. Schucher, supra note 137 at 392–94, 403–06 (providing a very succinct and helpful discussion of the significance of Meiorin in the development of human rights law).

E Shilton 425

tribunals will now impose a “higher evidentiary standard and a more individualized risk assessment” on insurance companies seeking to justify discriminatory practices.141 While they acknowledge that the Court had previously applied different justificatory standards to employment and non-employment cases, they argue that Meiorin’s synthesis of accommodation requirements with statutory defences of reasonableness and bona fides is now the law for all human rights purposes in Canada, making Zurich no longer binding authority, even for insurance cases.142

Lemmens and Thiery are almost certainly too optimistic about the probability that the Court will apply the Meiorin standard of individual assessment in all discrimination contexts. Their article, published in 2007, did not anticipate the Court’s 2008 decision in the Potash case.143 In that case, the Court explicitly refused to apply Meiorin in interpreting the exemption from age discrimination prohibitions in the New Brunswick Human Rights Act for mandatory retirement policies established by “any bona fide retirement or pension plan”.144 As the majority judges in Potash saw it, the application of human rights protection to pensions requires “different analytic frameworks” than similar protection for employment rights.145 In the pension context, they refused to demand a demonstration that mandatory retirement was required for the operation of the pension plan, or even that it had to be reasonably related to the operation of the plan. Instead, they held that a pension plan was bona fide if it was “a legitimate plan, adopted in good faith and not for the purpose of defeating protected rights”—a test that would exempt almost any registered

141. Lemmens & Thiery, supra note 138 at paras 54, 66–73, 85. 142. Ibid at para 64. In reaching this conclusion, Lemmens & Thiery

understandably relied on the Supreme Court’s observation in British Columbia (Superintendent of Motor Vehicles) v British Columbia (Council of Human Rights), [1999] 3 SCR 868 at para 19 that Meiorin applies to all discrimination claims. It is clear from Potash Corp of Saskatchewan Inc v Scott, 2008 SCC 45 at paras 14–20, [2008] 2 SCR 604, that the Court has now changed its mind. 143. Ibid. 144. RSNB 2011, c 171, s 4(6). 145. Potash, supra note 142 at paras 26–28.

426 (2012) 37:2 Queen’s LJ

plan.146 Chief Justice McLachlin, author of the Meiorin decision, dissented only in part. She would have insisted that the employer show a rational connection between the pension plan and its mandatory retirement policy, and establish that the policy was reasonably necessary to the plan’s sustainability.147 Importantly, however, she too found Meiorin’s requirement of individual assessment to be inappropriate in the pension context. As she saw it, “[p]ension plans are premised on treating groups of people in similar fashion. An individual approach belies this premise”.148

Like pensions, insurance is a group concept, a reality the Court duly noted twenty years ago in Zurich when it observed that “the insurance context is different from the employment context”.149 In Potash, the Court made it clear that Meiorin does not demand individualized assessment where that approach would be inconsistent with the inherent nature of the instrument at issue. It is predictable, therefore, that the Court will not imply a requirement of individualized risk assessment into the statutory human rights obligations of insurers. It need not do so, however, in order to find that sex-based annuity pricing constitutes unlawful discrimination. There is no credible evidence of any causal connection between sex and longevity. Although there is clearly a statistical relationship, it has become increasingly difficult to defend the stability of that relationship and its reliability as a predictor of important differences in life

146. Ibid at para 41. While the Court placed some emphasis on the fact that the

New Brunswick statutory language requires only bona fides and not reasonableness (ibid at para 31), the majority’s unwillingness to import a requirement of rational connection cannot turn on this distinction, since in the context of employment-related requirements, it is willing to read a reasonableness component into the term “bona fide” alone. 147. Ibid at para 57. 148. Ibid at para 78. Lemmens & Thiery, supra note 138, likewise did not

anticipate the court’s watering-down of Meiorin’s accommodation requirement even in the employment context. See McGill University Health Centre v Syndicat des employés de l’Hôpital général de Montréal, 2007 SCC 4 at para 20, [2007] 1 SCR 161; Hydro-Québec v Syndicat des employés de de techniques professionnelles et de bureau d’Hydro-Québec, 2008 SCC 43 at paras 14–22, [2008] 2 SCR 561. This development is discussed in Schucher, supra note 137 at 407. 149. Zurich, supra note 126 at para 22.

E Shilton 427

expectancy.150 SBMTs can now be readily replaced with unisex factors already in use in the industry. If, as I have argued above, SBMTs would likely have failed a straightforward application of the Zurich test back in 1992, subsequent legal and practical developments have only strengthened the case against them.

Even if the statutory Zurich test was found to be capacious enough to shelter SBMTs, I would still argue that their use is unlawful in Canada because compliance with statutory standards is not the only issue. The ultimate question for Canadian law is whether legislative provisions that may sanction their use of SBMTs conform to the equality rights protections under section 15 of the Canadian Charter of Rights and Freedoms.151 Under the current state of Canadian equality law, the answer to this question is not entirely straightforward. Since the earliest days of the Charter, the Supreme Court of Canada has adopted a commitment to substantive rather than formal equality, making it clear that a “mere distinction” based on a prohibited ground is not enough to ground a violation of section 15.152 For equality seekers, this approach has the important advantage of focusing the debate on the impact of an impugned law rather than on its form.153 However, the approach creates some uncertainties for challenges to laws that create or allow direct distinctions based on protected grounds. It means that a Canadian court confronted

150. See generally supra note 25. 151. See Potash, supra note 142 at para 4 (Abella J, who authored the majority

judgment, is at pains to point out that there was no Charter challenge in that case). Likewise, no Charter issues were raised in Zurich. 152. A useful capsule summary of the development of the Supreme Court of

Canada’s equality jurisprudence is found in “In Pursuit of Substantive Equality”, the introduction to Fay Faraday, Margaret Denike & M Kate Stephenson, eds, Making Equality Rights Real: Securing Substantive Equality under the Charter (Toronto: Irwin Law, 2006). See also Beverley Baines, “Equality, Comparison, Discrimination, Status” in ibid at 73; Denise G Réaume, “Discrimination and Dignity” in ibid at 123. 153. It also does much more than that, of course, although there is vigorous

academic debate over exactly what it does and whether, as it is currently applied by the Supreme Court of Canada, it has been beneficial for equality seekers. See Baines, supra note 152; Réaume, supra note 152; Sheila McIntyre, “Answering the Siren Call of Abstract Formalism with the Subjects and Verbs of Domination” in Faraday, Denike & Stephenson, supra note 152 at 99.

428 (2012) 37:2 Queen’s LJ

with a challenge to SBMTs will not take the strict categorical approach to the construction of risk categories reflected in the US decisions in Norris and Manhart. Rather than holding that it is unlawful per se to use sex as a risk classifier,154 a Canadian court will, quite properly, conduct an impact inquiry.

The analysis in Part II of this paper shows that SBMTs negatively affect women’s retirement income. Mere negative effect may not be sufficient, however, to ground a violation of section 15 of the Charter. Canadian courts have systematically demanded proof of something more than that. The key question is whether an impact analysis of the legal framework which supports SBMTs provides the required “something more”. The “something more” test is not an easy one to articulate.155 For several years, pursuant to the Law test,156 the Supreme Court required not only that a distinction harmed a rights claimant, but also that it violated the claimant’s human dignity. More recently, the Court appears to have abandoned its preoccupation with dignity.157 It now variously describes the search for a violation of section 15 as a search for a distinction that “in purpose or effect, perpetuates prejudice and disadvantage to members of a group on

154. Manhart and Norris were not constitutional decisions. If they had been,

they would have taken the court down a much more complex path, and one that would not have been likely to produce a decision in favour of the plaintiffs. See Owen M Fiss, “Groups and the Equal Protection Clause” (1976) 5:2 Phil & Publ Aff 107 (which predates these decisions but nevertheless provides an extremely useful discussion of how the US Supreme Court approaches equal protection issues). 155. The Court has not described its test(s) as “something more” tests; the

phrase is my own. The requirement for “something more” is, however, very clear in the Court’s articulation of its tests. 156. Law v Canada (Minister of Employment and Immigration), [1999] 1 SCR 497

at paras 4–10. 157. R v Kapp, 2008 SCC 41 at paras 21–25, [2008] 2 SCR 48. The dignity test

was intensely criticized by feminists and other commentators, a fact reflected in the unusually long list of law review articles cited in the Kapp decision. The dignity test also had its proponents; for insight into the feminist debate around the dignity test, see Faraday, Denike & Stephenson, supra note 152 (in particular, see the articles by Denise G Réaume, Sheila McIntyre, Fiona Sampson, and Gwen Brodsky & Shelagh Day).

E Shilton 429

the basis of personal characteristics within section 15(1)”.158 In addition to prejudice and disadvantage, the Court is also concerned with distinctions that “impose disadvantage on the basis of stereotyping”.159

Insurers argue that because SBMTs are generated “scientifically” by statistical analysis, they bear no kinship to distinctions based on prejudice or stereotypes. Opponents of SBMTs are not likely to concede this point; while statistical categories may not originate in prejudice, they are prototypical stereotypes, since they slot individuals based on categorical rather than individual characteristics.160 Even if insurers can persuade courts that SBMTs do not perpetuate stereotypes as that term is understood in the human rights context, they will have a hard time persuading them that SBMTs do not perpetuate disadvantage. Insurers will no doubt argue, as they have in the past, that SBMTs do not create disadvantage but simply reflect relevant sex-based longevity differences that exist in the real world. However, this argument is nothing more than a version of the “actuarial equality” argument dressed up in Charter language. As I have already argued in Part I, that argument is transparently circular. Longevity itself is not sex-based, and statistical sex differences in longevity are relevant for insurance purposes only if it is lawful both to measure longevity on the basis of sex and to use those measurements to shape insurance vehicles.

In any event, for purposes of a section 15 analysis it has never been necessary to show that the impugned laws create the disadvantage which is at the root of the claim. As the Supreme Court of Canada said in Withler, “[p]erpetuation of disadvantage typically occurs when the law treats a historically disadvantaged

158. Withler v Canada (AG), 2011 SCC 12 at para 35, [2011] 1 SCR 396.

Schucher, supra note 137 at 397–98 argues, correctly in my view, that a focus on the harms of stereotyping and prejudice is linked to an individual rights perspective on equality, while a focus on the harms of perpetuating disadvantage maps onto a group-based perspective. The casualness with which the Court mingles group and individual rights perspectives when enunciating the “something more” test suggests that it does not appreciate this distinction. 159. Kapp, supra note 157 at para 25. 160. See Zurich, supra note 126 at para 89, L’Heureux-Dubé J, dissenting.

430 (2012) 37:2 Queen’s LJ

group in a way that exacerbates the situation of the group”.161 Without venturing an exhaustive definition of “exacerbate”, I would argue that to show a violation of section 15, it is necessary to show only that the impugned law plays a role in entrenching an existing disadvantage. Certainly it is enough to show that the law makes an existing disadvantage worse. The analysis in this paper firmly links SBMTs in Canada to the retirement income inequality of Canadian women, and demonstrates how any expanded role for SBMTs will increase women’s already disadvantaged position in respect of retirement income. It also demonstrates that although SBMTs are artefacts of the market, they continue to thrive because of how legal rules, and the interpretation of those rules, regulate their use. It is therefore likely, in my view, that a Canadian court will find common ground with the ECJ in Test-Achats and hold that the law which sustains the use of SBMTs violates equality guarantees.

Conclusion

In this paper, I have argued that sex-based annuity pricing violates Canadian human rights laws and constitutional equality guarantees. The legal question is an interesting one, but for equality seekers and policy-makers, the practical questions are more important. Will banning the use of SBMTs on annuity markets make a difference for the retirement income equality of Canadian women? In other words, if we change our rules to require insurers to use gender-neutral risk categories, as Europe has done, will there be a positive impact on the gender pension gap? The answer is almost certainly yes, although the direct effect is likely to be a small one.

More important to the overall gender pension equality project may be the broader lessons reflected in the legal history of SBMTs. I have offered this account as a case study; what can we learn from it? First, we have seen that legal rules have direct effects on distributive outcomes. As obvious as this point may be, it bears emphasis in the context of a policy debate in which

161. Withler, supra note 158 at para 35.

E Shilton 431

opponents of expanded public pensions seek to persuade us that deregulation is an unqualified public good. This paper has shown that the legal rules governing employment pensions result in a different distribution of longevity risk than the legal rules that govern annuities bought and sold on the open market. It has also shown that different individuals and groups benefit from these different sets of legal rules, and that the current differences in the rules are costing Canadian women over $1 billion a year.

The second and related lesson, evident both from the evolution of SBMTs and the distributive role they have continued to play, is that if we allow retirement income instruments and the legal rules that govern them to be shaped by markets, they will reflect market norms and values. We cannot count on the market’s distributive principles to achieve equality for us; as we have seen, market distribution tends along quite different vectors. Current market-made tools for distributing retirement risks require women to pay more than men for protection from longevity risk. The fact that this leaves women with less retirement income to live on from year to year is not just an unfortunate by-product of market logic; perversely, it is integral to that logic, and imposes real costs on Canadian women. It is foreseeable that market mechanisms will have perverse effects on annuity pricing even in the absence of SBMTs; some economists predict that if unisex mortality tables are imposed by law, market mechanisms may peg unisex prices at levels close to or even as high as current female prices.162 It is also predictable that if market actors replace gender with disaggregated risk factors more plausibly linked to longevity, the result may simply shift

162. See e.g. Samuel A Rea Jr, “The Market Response to the Elimination of Sex-

based Annuities” (1987) 54 Southern Economic Journal 55; Equal Opportunities Commission, An Analysis of Unisex Annuity Rates by Chris Curry & Alison O’Connell (London, UK: Pensions Policy Institute, 2004); Amy Finkelstein, James Poterba & Casey Rothschild, “Redistribution by Insurance Market Regulation: Analyzing a Ban on Gender-Based Retirement Annuities” (2009) 91 Journal of Financial Economics 38. These predictions anticipate an increase in adverse selection problems and in industry costs associated with more individualized risk assessment. They also reflect the uncertainty associated with the abandonment of tried-and-true statistical measures, an uncertainty to which the industry will likely respond by increasing capital reserves.

432 (2012) 37:2 Queen’s LJ

inequalities from “women” as a category to other disadvantaged groups, or to more disadvantaged sub-groups of women, thereby producing an overall result at least as unattractive from a social policy perspective as the current explicitly gendered result.163 It is also obvious that whatever impact the elimination of SBMTs may have on the relatively small percentage of the retirement income gap for which it is directly responsible, it will not touch the much larger percentage—the portion attributable not to direct sex discrimination, but to the practical problem that women have less money than men (less “pension wealth”) for reasons that are deeply imbedded in the social and economic organization of labour markets and families.

This brings us to the third important lesson gleaned from the history of SBMTs: that approaches to addressing the risks of retirement which focus on individual rather than collective solutions are likely to exacerbate inequality rather than alleviate it. The US legal battle over SBMTs has been constructed as a clash between individual rights and group rights, in which the former have prevailed. The European approach is less transparently individualistic, but it still turns on a narrow conception of equality that treats “likes” alike, and permits the “unlike” to be treated as different. The principal evil of discrimination is seen as

163. Classification factors that are related to health status have been suggested as

possible substitutes if SBMTs are prohibited. These factors include smoking behaviour, obesity, genetic predisposition to certain serious diseases, health risks that emerge later in life (causally connected with longevity), income level/social class (statistically associated with longevity) and geographic location (associated with both socio-economic status and environmental exposure). See generally Curry & O’Connell, supra note 162 at 7; Ontario Human Rights Commission, Human Rights Issues in Insurance, Discussion Paper (Toronto: Policy and Education Branch, 1999); Ontario Human Rights Commission, Human Rights Issues in Insurance, Consultation Report (Toronto: Policy and Education Branch, 2001). Some of these factors might shift the higher costs of greater longevity to groups in a better position to bear it; for example, higher socio-economic status is normally associated with higher longevity. Others might have the opposite effect. With respect to disability and property insurance, there is a vigorous ongoing debate about “fair” insurance categories, with scholars and activists underscoring the negative impact on welfare outcomes if markets are allowed to dictate the availability and pricing of insurance for important social goods. See e.g. Lemmens, supra note 8.

E Shilton 433

the ascription of group characteristics to those who do not possess them as individuals. In both the US and in Europe, the “actuarial value” argument stumbled at least in part because of the absence of any proven causal link between sex and longevity: not all women share the characteristic of longevity. Implicitly, individual rights analysis endorses the proposition that sex-based risk categories might be justified if longevity were indeed causally linked to sex: if all women lived longer than all men.164 From a social policy perspective, however, the impact of SBMTs on women’s standard of living in retirement would be just as negative—elderly women would still have fewer resources to live on in retirement than elderly men.

In fact, it is not grouping women with other women that makes SBMTs an instrument of inequality; it is that SBMTs draw group boundaries with a view to avoiding any meaningful pooling of longevity risks. A strict individual rights approach to retirement income, even if it were not explicitly sex-based, would only make this problem worse. In a context where women’s inequalities are deeply woven into the fabric of socio-economic structures, which are on their face gender-neutral, treating women as individuals leaves them to their own (and substantively unequal) resources. This leaves them more exposed than men not only to longevity risk, but also to the host of other risks that flow from their unequal status.

There is more than a little irony in the fact that in the contest between group rights and individual rights, the insurance industry has been cast as the champion of group rights.165 As we have seen, the insurance notion of fairness is ultimately individual rather than collective in orientation; it admits no notion of social solidarity. Private insurance markets favour narrow risk categories that are carefully designed to protect some individuals from being called upon to subsidize other individuals who are more exposed to the hazards insured against. The notion of

164. Both the US and European case law are ambiguous on this point. 165. The exemption from human rights principles that Canadian law currently

gives to the insurance industry has been justified by the perceived incompatibility between the individual nature of human rights protection and the group nature of insurance. See Zurich, supra note 126 at para 17.

434 (2012) 37:2 Queen’s LJ

insurance, bare of its market context, does not dictate this ungenerous approach to risk categories. In fact, as Wouter Wils points out, “[t]he contrary may arguably be the case, as the essence of insurance is the pooling of risk, not discrimination between risks”.166 This lesson was well-learned by the architects of the post-war welfare state, who gave social insurance schemes an important role in welfare design—schemes like the CPP/QPP, which distribute the risks of welfare loss in retirement broadly across the workforce. Current policy makers are in danger of forgetting the lesson, besieged as they are by powerful voices arguing that interference with markets will upset some naturally ordained distributive balance in the allocation of social and economic resources.

Within the current legal framework that governs annuity pricing in Canada, Canadian women will see their position worsened if policy-makers choose retirement income reform options that enhance the role of annuities sold on the open market. Changes to the legal rules will benefit women, but the extent of that benefit will depend on the extent to which policy makers understand the broader lessons to be learned from the history of SBMTs: that legal rules play an important role in shaping distributive outcomes, that legal rules shaped by the prevailing logic of the market work against equality-driven outcomes, and that retirement income instruments based on assigning the welfare risks of retirement to individuals are less likely to be effective at closing the pension gap than approaches which treat those risks as social risks. Policy-makers who understand this and who value equality will choose an active role for the state in the design of future retirement income vehicles, providing an opportunity167 to pool and distribute retirement income risks among all Canadians. Women, who currently bear a

166. Wils, supra note 14 at 451. 167. It is important to caution that state involvement in the design of retirement

income instruments simply presents an opportunity to enhance equality. Equality does not flow automatically from state action. It is the specific design features of new retirement income instruments that will ultimately control their distributive outcomes.

E Shilton 435

disproportionate share of the risks of welfare loss in retirement, will be the winners.